define sample

The analyses below are based on the sample of subset subjects.

For subset subjects, subjects (n = 5) were excluded according to the preregistration.
Preregistration: https://doi.org/10.17605/OSF.IO/UYJVA
These subjects had less than 70% of audio epochs in any condition after artifact rejection that excluded eye blinks.

For all subjects, all subjects were included, and only extreme artifacts were rejected (i.e., eye blinks were ignored). This is reasonable because ASSRs occur at a much higher frequency than eye blinks.

tmpaEEG=read.csv("data_audio_signal_noise.csv", sep="\t",header=TRUE, dec=".");
#head(tmpaEEG)

# The data in each 3-min block were analyzed by minute. For each minute, mean amp (and itc) were computed and then averaged across the four blocks of load.
tmpa2EEG=read.csv("data_audio_signal_noise_by_minute.csv", sep="\t",header=TRUE, dec=".");
#head(tmpa3EEG)

# Analysis of visual P3, as in our MMN paper (2019)
tmpvEEG=read.csv("data_visual_P3.csv", sep="\t",header=TRUE, dec=".");
#head(tmpwmc)

tmpbeh=read.csv("data_logfiles.txt", sep="\t",header=TRUE, dec=".");
#head(tmpbeh)

tmpwmc=read.csv("data_wmc.txt", sep="\t",header=TRUE, dec=".");
#head(tmpwmc)

D=merge(tmpbeh, tmpwmc, by = "fp")
D=merge(D, tmpaEEG, by = "fp")
D=merge(D, tmpa2EEG, by = "fp")
D=merge(D, tmpvEEG, by = "fp")
rm(list=ls(pattern="tmp"))

if (any(as.numeric(table(D$fp))!=1)){
  stop("Subject numbers are not unique!")
  }

# maybe exclude subjects? see above
if (nowsample == "subset") {
  excludefps = c(3, 20, 24, 35, 40)
  indx = D$fp %in% excludefps
  D = D[!indx,]}
  #table(D$fp)

Sample size: N = 38.

quick data check

Sanity check for behavioral performance (across low and high load):

min number of hits (could be 0) : 147
max number of hits (could be 288) : 288
min number of false alarms (could be 0) : 0
max number of false alarms (could be 892) : 95

Note:
Hits are correct responses between 200 and 1000 ms after target onset (defined in ASSR_study1_process_beh.R). This criterion was not preregistered, but it makes sense.
False alarms are responses at other times: 1152 + 7*4 (initial trials) - 288 (trial after a target) = 892.

# add 0.5 and 1 to avoid ceiling/floor effects
D$HRlocorr = (D$nHRlo + 0.5)/(288+1)
D$HRhicorr = (D$nHRhi + 0.5)/(288+1)
# 72*4 = 288
D$FAlocorr = (D$nFAlo + 0.5)/(1180+1-288)
D$FAhicorr = (D$nFAhi + 0.5)/(1180+1-288)
# 288*4 = 1152 + 7*4 (initial trials) = 1180
# do not count the next trial after a target (because this is considered to be a valid response)
# Dprime 
D$dprlo <- qnorm(D$HRlocorr) - qnorm(D$FAlocorr)
D$dprhi <- qnorm(D$HRhicorr) - qnorm(D$FAhicorr)
# no need for criterion
#tmp$criterion <- - (1/2) * (qnorm(tmp$aware_crit_corrected) + qnorm(tmp$aware_catch_corrected))
D$dprloMhi = D$dprlo - D$dprhi
# RT to hits
D$meanRTloMhi = D$meanRTlo_ms - D$meanRThi_ms

# the numbers 1 to 4 refer to different blocks

# amp SNR raw (this measure was preregistered but turned out useless, see below)
D$AmpSNRlo = rowMeans(cbind(D$AmpSNRlo1,D$AmpSNRlo2,D$AmpSNRlo3,D$AmpSNRlo4))
D$AmpSNRhi = rowMeans(cbind(D$AmpSNRhi1,D$AmpSNRhi2,D$AmpSNRhi3,D$AmpSNRhi4))
D$AmpSNRloMhi = D$AmpSNRlo - D$AmpSNRhi
D$AmpSNRlo1Mhi1 = D$AmpSNRlo1 - D$AmpSNRhi1
D$AmpSNRlo2Mhi2 = D$AmpSNRlo2 - D$AmpSNRhi2
D$AmpSNRlo3Mhi3 = D$AmpSNRlo3 - D$AmpSNRhi3
D$AmpSNRlo4Mhi4 = D$AmpSNRlo4 - D$AmpSNRhi4

# amp S raw (signal)
D$AmpSlo = rowMeans(cbind(D$AmpSlo1,D$AmpSlo2,D$AmpSlo3,D$AmpSlo4))
D$AmpShi = rowMeans(cbind(D$AmpShi1,D$AmpShi2,D$AmpShi3,D$AmpShi4))
D$AmpSloMhi = D$AmpSlo - D$AmpShi
D$AmpSlo1Mhi1 = D$AmpSlo1 - D$AmpShi1
D$AmpSlo2Mhi2 = D$AmpSlo2 - D$AmpShi2
D$AmpSlo3Mhi3 = D$AmpSlo3 - D$AmpShi3
D$AmpSlo4Mhi4 = D$AmpSlo4 - D$AmpShi4

# amp N raw (noise)
D$AmpNlo = rowMeans(cbind(D$AmpNlo1,D$AmpNlo2,D$AmpNlo3,D$AmpNlo4))
D$AmpNhi = rowMeans(cbind(D$AmpNhi1,D$AmpNhi2,D$AmpNhi3,D$AmpNhi4))
D$AmpNloMhi = D$AmpNlo - D$AmpNhi
D$AmpNlo1Mhi1 = D$AmpNlo1 - D$AmpNhi1
D$AmpNlo2Mhi2 = D$AmpNlo2 - D$AmpNhi2
D$AmpNlo3Mhi3 = D$AmpNlo3 - D$AmpNhi3
D$AmpNlo4Mhi4 = D$AmpNlo4 - D$AmpNhi4

# amp S minus N (nice because it compares S with N)
D$AmpSmNlo = rowMeans(cbind(D$AmpSmNlo1,D$AmpSmNlo2,D$AmpSmNlo3,D$AmpSmNlo4))
D$AmpSmNhi = rowMeans(cbind(D$AmpSmNhi1,D$AmpSmNhi2,D$AmpSmNhi3,D$AmpSmNhi4))
D$AmpSmNloMhi = D$AmpSmNlo - D$AmpSmNhi
D$AmpSmNlo1Mhi1 = D$AmpSmNlo1 - D$AmpSmNhi1
D$AmpSmNlo2Mhi2 = D$AmpSmNlo2 - D$AmpSmNhi2
D$AmpSmNlo3Mhi3 = D$AmpSmNlo3 - D$AmpSmNhi3
D$AmpSmNlo4Mhi4 = D$AmpSmNlo4 - D$AmpSmNhi4

# the numbers 1 to 4 refer to different blocks

# itc SNR raw (this measure was preregistered but turned out useless, see below)
D$ItcSNRlo = rowMeans(cbind(D$ItcSNRlo1,D$ItcSNRlo2,D$ItcSNRlo3,D$ItcSNRlo4))
D$ItcSNRhi = rowMeans(cbind(D$ItcSNRhi1,D$ItcSNRhi2,D$ItcSNRhi3,D$ItcSNRhi4))
D$ItcSNRloMhi = D$ItcSNRlo - D$ItcSNRhi
D$ItcSNRlo1Mhi1 = D$ItcSNRlo1 - D$ItcSNRhi1
D$ItcSNRlo2Mhi2 = D$ItcSNRlo2 - D$ItcSNRhi2
D$ItcSNRlo3Mhi3 = D$ItcSNRlo3 - D$ItcSNRhi3
D$ItcSNRlo4Mhi4 = D$ItcSNRlo4 - D$ItcSNRhi4

# itc S raw (signal)
D$ItcSlo = rowMeans(cbind(D$ItcSlo1,D$ItcSlo2,D$ItcSlo3,D$ItcSlo4))
D$ItcShi = rowMeans(cbind(D$ItcShi1,D$ItcShi2,D$ItcShi3,D$ItcShi4))
D$ItcSloMhi = D$ItcSlo - D$ItcShi
D$ItcSlo1Mhi1 = D$ItcSlo1 - D$ItcShi1
D$ItcSlo2Mhi2 = D$ItcSlo2 - D$ItcShi2
D$ItcSlo3Mhi3 = D$ItcSlo3 - D$ItcShi3
D$ItcSlo4Mhi4 = D$ItcSlo4 - D$ItcShi4

# itc N raw (noise)
D$ItcNlo = rowMeans(cbind(D$ItcNlo1,D$ItcNlo2,D$ItcNlo3,D$ItcNlo4))
D$ItcNhi = rowMeans(cbind(D$ItcNhi1,D$ItcNhi2,D$ItcNhi3,D$ItcNhi4))
D$ItcNloMhi = D$ItcNlo - D$ItcNhi
D$ItcNlo1Mhi1 = D$ItcNlo1 - D$ItcNhi1
D$ItcNlo2Mhi2 = D$ItcNlo2 - D$ItcNhi2
D$ItcNlo3Mhi3 = D$ItcNlo3 - D$ItcNhi3
D$ItcNlo4Mhi4 = D$ItcNlo4 - D$ItcNhi4

# itc S minus N (nice because it compares S with N)
D$ItcSmNlo = rowMeans(cbind(D$ItcSmNlo1,D$ItcSmNlo2,D$ItcSmNlo3,D$ItcSmNlo4))
D$ItcSmNhi = rowMeans(cbind(D$ItcSmNhi1,D$ItcSmNhi2,D$ItcSmNhi3,D$ItcSmNhi4))
D$ItcSmNloMhi = D$ItcSmNlo - D$ItcSmNhi
D$ItcSmNlo1Mhi1 = D$ItcSmNlo1 - D$ItcSmNhi1
D$ItcSmNlo2Mhi2 = D$ItcSmNlo2 - D$ItcSmNhi2
D$ItcSmNlo3Mhi3 = D$ItcSmNlo3 - D$ItcSmNhi3
D$ItcSmNlo4Mhi4 = D$ItcSmNlo4 - D$ItcSmNhi4

# visual P3 is difference between targets and nontargets
# the numbers 1 to 4 refer to different blocks
D$vP3lo1 = D$vP3_Tarlo1- D$vP3_Nonlo1
D$vP3lo2 = D$vP3_Tarlo2- D$vP3_Nonlo2
D$vP3lo3 = D$vP3_Tarlo3- D$vP3_Nonlo3
D$vP3lo4 = D$vP3_Tarlo4- D$vP3_Nonlo4

D$vP3hi1 = D$vP3_Tarhi1- D$vP3_Nonhi1
D$vP3hi2 = D$vP3_Tarhi2- D$vP3_Nonhi2
D$vP3hi3 = D$vP3_Tarhi3- D$vP3_Nonhi3
D$vP3hi4 = D$vP3_Tarhi4- D$vP3_Nonhi4

D$vP3lo = rowMeans(cbind(D$vP3lo1, D$vP3lo2, D$vP3lo3, D$vP3lo4))
D$vP3hi = rowMeans(cbind(D$vP3hi1, D$vP3hi2, D$vP3hi3, D$vP3hi4))
D$vP3loMhi = D$vP3lo - D$vP3hi

hit rates

Simple hit rates (%) were as follows:

Low load = 95.51
High load = 80.77

tables

behavior

dpr: signal detection index d´
RT: reaction time to hits (in ms)
pcu: partial credit unit score of working memory capacity
LL and UL refer to the 95% confidence interval (from two-tailed t tests).

Dtmp = subset(D, select=c(dprlo, dprhi, dprloMhi, meanRTlo_ms, meanRThi_ms, meanRTloMhi, pcu))
RmCI = tableCIs(Dtmp)
RmCI %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Mean	LL	UL	N
dprlo	4.336	4.099	4.574	38
dprhi	2.641	2.437	2.846	38
dprloMhi	1.695	1.500	1.890	38
meanRTlo_ms	376.805	363.479	390.130	38
meanRThi_ms	504.650	489.222	520.079	38
meanRTloMhi	-127.846	-138.273	-117.419	38
pcu	0.731	0.682	0.780	38

EEG

Amp: amplitude
Itc: intertrial coherence
SNR: signal-to-noise ratio
S: signal
N: noise
SmN: signal minus noise
vP3: visual P3 to targets (vs. nontargets)
LL and UL refer to the 95% confidence interval (from two-tailed t tests).

Dtmp = subset(D, select=c(AmpSNRlo, AmpSNRhi, AmpSNRloMhi,
                           AmpSlo, AmpShi, AmpSloMhi,
                           AmpNlo, AmpNhi, AmpNloMhi,
                           AmpSmNlo, AmpSmNhi, AmpSmNloMhi,
                           ItcSNRlo, ItcSNRhi, ItcSNRloMhi,
                           ItcSlo, ItcShi, ItcSloMhi,
                           ItcNlo, ItcNhi, ItcNloMhi,
                           ItcSmNlo, ItcSmNhi, ItcSmNloMhi,
                           vP3lo, vP3hi, vP3loMhi))
RmCI = tableCIs(Dtmp)
RmCI %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Mean	LL	UL	N
AmpSNRlo	5.774	4.882	6.666	38
AmpSNRhi	5.473	4.585	6.362	38
AmpSNRloMhi	0.301	-0.184	0.786	38
AmpSlo	0.239	0.208	0.269	38
AmpShi	0.236	0.206	0.267	38
AmpSloMhi	0.002	-0.008	0.013	38
AmpNlo	0.045	0.041	0.049	38
AmpNhi	0.048	0.044	0.053	38
AmpNloMhi	-0.003	-0.006	-0.001	38
AmpSmNlo	0.194	0.163	0.224	38
AmpSmNhi	0.188	0.157	0.219	38
AmpSmNloMhi	0.006	-0.006	0.017	38
ItcSNRlo	5.006	4.419	5.594	38
ItcSNRhi	4.744	4.121	5.366	38
ItcSNRloMhi	0.262	-0.082	0.607	38
ItcSlo	0.397	0.350	0.444	38
ItcShi	0.379	0.331	0.428	38
ItcSloMhi	0.017	-0.003	0.038	38
ItcNlo	0.080	0.079	0.082	38
ItcNhi	0.082	0.080	0.084	38
ItcNloMhi	-0.002	-0.004	0.001	38
ItcSmNlo	0.317	0.270	0.364	38
ItcSmNhi	0.297	0.249	0.346	38
ItcSmNloMhi	0.019	-0.002	0.041	38
vP3lo	8.538	6.604	10.472	38
vP3hi	2.501	0.980	4.021	38
vP3loMhi	6.037	4.692	7.383	38

combined

dpr: signal detection index d´
RT: reaction time to hits (in ms)
pcu: partial credit unit score of working memory capacity
Amp: amplitude
Itc: intertrial coherence
SNR: signal-to-noise ratio
S: signal
N: noise
SmN: signal minus noise
vP3: visual P3 to targets (vs. nontargets)
LL and UL refer to the 95% confidence interval (from two-tailed t tests).

# Dtmp = subset(D, select=c(dprlo, dprloMhi, 
#                           meanRTlo_ms, meanRTloMhi,
#                           pcu,
#                           AmpSNRlo, AmpSNRloMhi,
#                           AmpSlo, AmpSloMhi,
#                           AmpSmNlo, AmpSmNloMhi,
#                           ItcSNRlo, ItcSNRloMhi,
#                           ItcSlo, ItcSloMhi,
#                           ItcSmNlo, ItcSmNloMhi,
#                           vP3lo, vP3loMhi))
Dtmp = subset(D, select=c(dprlo, dprhi, dprloMhi, 
                          meanRTlo_ms, meanRThi_ms, meanRTloMhi,
                          pcu,
                          AmpSmNlo, AmpSmNhi, AmpSmNloMhi,
                          ItcSmNlo, ItcSmNhi, ItcSmNloMhi,
                          vP3lo, vP3hi, vP3loMhi))
RmCI = tableCIs(Dtmp)
RmCI %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Mean	LL	UL	N
dprlo	4.336	4.099	4.574	38
dprhi	2.641	2.437	2.846	38
dprloMhi	1.695	1.500	1.890	38
meanRTlo_ms	376.805	363.479	390.130	38
meanRThi_ms	504.650	489.222	520.079	38
meanRTloMhi	-127.846	-138.273	-117.419	38
pcu	0.731	0.682	0.780	38
AmpSmNlo	0.194	0.163	0.224	38
AmpSmNhi	0.188	0.157	0.219	38
AmpSmNloMhi	0.006	-0.006	0.017	38
ItcSmNlo	0.317	0.270	0.364	38
ItcSmNhi	0.297	0.249	0.346	38
ItcSmNloMhi	0.019	-0.002	0.041	38
vP3lo	8.538	6.604	10.472	38
vP3hi	2.501	0.980	4.021	38
vP3loMhi	6.037	4.692	7.383	38

write.table(RmCI, file = file.path(dir_fig, paste0("table_meansCI_",nowsample,"_subjects.tsv")), row.names = F, dec = ".", sep = "\t")

Bayes factors for load

Bayes factor (BF) analyses from Bayesian one-sample t tests of difference scores (low minus high load).
The BFs were computed with Aladins R script.
https://doi.org/10.17045/sthlmuni.4981154.v3

The BF01 uses uniform H1 models with different lower limits (LL) and upper limits (UL).
BF01 is the evidence for the null hypothesis relative to the alternative hypothesis.
If BF01 > 3, this is evidence for the null.
If BF01 < 1/3, this is evidence against the null.

Dtmp2 = subset(D, select=c(AmpSNRloMhi, AmpSloMhi, AmpNloMhi, AmpSmNloMhi, 
                          ItcSNRloMhi, ItcSloMhi, ItcNloMhi, ItcSmNloMhi))

LL = -1
UL = 1
BF10_Q1_U = numeric()
Vlabels = c("ampSNR", "ampS", "ampN", "ampSmN", "itcSNR", "itcS", "itcN", "itcSmN")
#Vlabels = c("ampSmN","itcSmN")
for (i in 1:length(Vlabels)) {
  tmp = Dtmp2[,i]
  meanobtained = mean(tmp)
  semobtained = sd(tmp)/sqrt(length(tmp))
  dfobtained = length(tmp)-1
  BF10_Q1_U[i] = BF_U(LL, UL, meanobtained, semobtained, dfobtained)
}
BF01_Q1_U = 1/BF10_Q1_U

LL = 0
UL = 1
BF10_Q2_U = numeric()
for (i in 1:length(Vlabels)) {
  tmp = Dtmp2[,i]
  meanobtained = mean(tmp)
  semobtained = sd(tmp)/sqrt(length(tmp))
  dfobtained = length(tmp)-1
  BF10_Q2_U[i] = BF_U(LL, UL, meanobtained, semobtained, dfobtained)
}
BF01_Q2_U = 1/BF10_Q2_U

LL = 0
UL = 0.2
BF10_Q3_U = numeric()
for (i in 1:length(Vlabels)) {
  tmp = Dtmp2[,i]
  meanobtained = mean(tmp)
  semobtained = sd(tmp)/sqrt(length(tmp))
  dfobtained = length(tmp)-1
  BF10_Q3_U[i] = BF_U(LL, UL, meanobtained, semobtained, dfobtained)
}
BF01_Q3_U = 1/BF10_Q3_U

RBF = cbind(BF01_Q1_U, BF01_Q2_U, BF01_Q3_U)
RBF = round(RBF, digits = 1)
RBF = data.frame(Vlabels, RBF)
colnames(RBF) = c('Variable','[-1, +1]','[0, +1]','[0, +0.2]')
RBF_sel = RBF[RBF$Variable %in% c("ampSmN","itcSmN"),]
RBF %>%
  kable(digits = 3) %>%
  add_header_above(c(" " = 1, "BF01" = 3)) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

	BF01
Variable	[-1, +1]	[0, +1]	[0, +0.2]
ampSNR	1.5	0.8	0.7
ampS	137.8	98.0	20.8
ampN	34.7	839.4	429.0
ampSmN	85.9	50.7	10.4
itcSNR	1.4	0.8	0.5
itcS	18.7	9.8	2.0
itcN	214.6	777.1	335.8
itcSmN	14.6	7.5	1.5

# save only SmN
write.table(RBF_sel, file = file.path(dir_fig, paste0("table_SmN_BF_",nowsample,"_subjects.tsv")), row.names = F, dec = ".", sep = "\t")

switch from SNR to SmN

explanation

Originally, we preregistered to compute the signal-to-noise ratio (SNR) of amplitudes to obtain a single measure that captures the strength of the signal relative to the noise. Amplitude SNR was calculated as the ratio between the amplitude at 40.96 Hz and the mean amplitude across 20 neighboring frequencies (ten on each side but omitting the two immediate neighbors on each side).

However, during data analysis, we realized that amplitude SNR was not the best measure of the differences between signal and noise. Because in our study, noise levels were very close to 0, amplitude SNR was extremely variable, even when signal and noise levels were almost identical. For example, consider the following realistic amplitude values: If for low load, signal amplitude = 0.24 µV and noise amplitude = 0.045 µV, then amplitude SNR = 5.33; and if for high load, signal = 0.24 µV and noise = 0.048 µV, then amplitude SNR = 5.0. Even though the amplitude is the same in low and high load for the signal, and almost the same for the noise (difference = −0.003 µV), the amplitude SNRs differ substantially (difference = 0.33). Thus, with small noise levels, tiny differences in noise levels yield substantial variability in amplitude SNR. As a consequence, results were inconclusive for SNR, even though results were conclusive for separate analyses of signal and noise (see above).

To obtain a single measure of the difference between signal and noise, we computed the amplitude difference of signal minus noise (SmN), defined as the amplitude at 40.96 Hz minus the mean amplitude across the neighboring 20 frequencies (ten on each side but omitting two immediate neighbors).

Although SmN as well as SNR are computed here, the manuscript reports only SmN.

SNR is noisy

Dtmp = subset(D, select=
                c(fp, AmpSNRlo, AmpSNRhi, AmpSNRloMhi, ItcSNRlo, ItcSNRhi, ItcSNRloMhi))
Dtmpsd <- Dtmp %>% summarize(AmpSNRloMhi = sd(AmpSNRloMhi, na.rm = TRUE),
                             ItcSNRloMhi = sd(ItcSNRloMhi, na.rm = TRUE))

As shown below, the scores vary greatly for the signal-to-noise ratio (SNR). The range of the differences between low and high load is huge for amplitude and ITC.

The difference scores between low and high load do not appear to be normally distributed (Shapiro-Wilks test):
AmpSNRloMhi: p = 0.146
ItcSNRloMhi: p = 0.014

Standard deviations:
AmpSNRloMhi: 1.476
ItcSNRloMhi: 1.048

plotoverlay(dm = Dtmp,lbl = "SNR: Means and 95% CIs for low minus high")

hist(Dtmp$AmpSNRloMhi, main = "Amplitude SNR low minus high", xlab = "Amplitude SNR low minus high")

hist(Dtmp$ItcSNRloMhi, main = "ITC SNR low minus high", xlab = "ITC SNR low minus high")

SmN is better

Dtmp = subset(D, select=
                c(fp, AmpSmNlo, AmpSmNhi, AmpSmNloMhi, ItcSmNlo, ItcSmNhi, ItcSmNloMhi))
Dtmpsd <- Dtmp %>% summarize(AmpSmNloMhi = sd(AmpSmNloMhi, na.rm = TRUE),
                             ItcSmNloMhi = sd(ItcSmNloMhi, na.rm = TRUE))

For signal-minus-noise (SmN), the difference scores are more reasonable. They appear to be normally distributed and have a narrower range.

Test of normal distribution of difference scores between low and high load (Shapiro-Wilks test):
AmpSmNloMhi: p = 0.863
ItcSmNloMhi: p = 0.099

Standard deviations:
AmpSmNloMhi: 0.036
ItcSmNloMhi: 0.065

plotoverlay(dm = Dtmp,lbl = "SmN: Means and 95% CIs for low minus high")

hist(Dtmp$AmpSmNloMhi, main = "Amplitude SmN low minus high", xlab = "Amplitude SmN low minus high")

hist(Dtmp$ItcSmNloMhi, main = "ITC SmN low minus high", xlab = "ITC SmN low minus high")

plot amp and ITC

Dtmp = subset(D, select=
                c(fp, AmpSmNlo, AmpSmNhi, AmpSmNloMhi))
colnames(Dtmp) <- c('fp','low', 'high', 'loMhi')
bp1 = plotoverlay(dm = Dtmp,lbl = "Amplitude SmN", dvy = "Mean amplitude (µV)")

Dtmp = subset(D, select=
                c(fp, ItcSmNlo, ItcSmNhi, ItcSmNloMhi))
colnames(Dtmp) <- c('fp','low', 'high', 'loMhi')
bp2 = plotoverlay(dm = Dtmp,lbl = "ITC SmN", dvy = "Mean ITC")
grid.arrange(bp1, bp2, ncol = 2)

bp3 = arrangeGrob(bp1, bp2, ncol = 2)
ggsave(file.path(dir_fig, paste0("figure_SmN_meansCI_amp_itc_",nowsample,"_subjects.pdf")), bp3, dpi = 600, width = 7, height = 4)
rm(bp1, bp2, bp3)

correlations with behavior

table

Dtmp = subset(D, select=c(AmpSNRloMhi, AmpSmNloMhi, 
                          ItcSNRloMhi, ItcSmNloMhi,
                          vP3loMhi, dprloMhi, meanRTloMhi))
cor(Dtmp) %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

	AmpSNRloMhi	AmpSmNloMhi	ItcSNRloMhi	ItcSmNloMhi	vP3loMhi	dprloMhi	meanRTloMhi
AmpSNRloMhi	1.000	0.790	0.820	0.865	0.138	-0.162	-0.189
AmpSmNloMhi	0.790	1.000	0.721	0.852	0.000	-0.067	-0.053
ItcSNRloMhi	0.820	0.721	1.000	0.916	0.056	0.001	-0.234
ItcSmNloMhi	0.865	0.852	0.916	1.000	0.111	-0.039	-0.237
vP3loMhi	0.138	0.000	0.056	0.111	1.000	0.337	-0.132
dprloMhi	-0.162	-0.067	0.001	-0.039	0.337	1.000	-0.199
meanRTloMhi	-0.189	-0.053	-0.234	-0.237	-0.132	-0.199	1.000

scatterplots

Dtmp = subset(D, select=c(AmpSmNloMhi, ItcSmNloMhi,
                          vP3loMhi, dprloMhi, meanRTloMhi))
pairs(Dtmp)

fishing

For the large sample, data exploration (fishing) suggests that the more d prime decreased from low to high, the larger the visual P3 decreased from low to high.

cor.test(D$dprloMhi, D$vP3loMhi)

## 
##  Pearson's product-moment correlation
## 
## data:  D$dprloMhi and D$vP3loMhi
## t = 2.1495, df = 36, p-value = 0.03839
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.01970551 0.59301054
## sample estimates:
##       cor 
## 0.3372643

correlations with wm capacity

Pearson correlations with working memory capacity.

table

LL and UL refer to the 95% confidence interval of the correlation (from one-tailed t tests).
The BFs are computed with the BayesFactor package in R.
The BF01 uses a flat prior (beta width = 1). This prior is the default in JASP.
Cred_LL and Cred_UL refer to the 95% credible interval of the correlation (from BayesFactor).

Dtmp = subset(D, select=
                c(fp, 
                 AmpSNRlo, AmpSlo, AmpSmNlo,
                 ItcSNRlo, ItcSlo, ItcSmNlo,
                 AmpSNRhi, AmpShi, AmpSmNhi,
                 ItcSNRhi, ItcShi, ItcSmNhi,
                 AmpSNRloMhi, AmpSloMhi, AmpSmNloMhi,
                 ItcSNRloMhi, ItcSloMhi, ItcSmNloMhi))

Vlabels = colnames(Dtmp)
Vlabels = Vlabels[-1]
Rcorr = matrix(nrow = length(Vlabels), ncol = 7)
Dtmp$pcu = D$pcu
Dtmp = na.omit(Dtmp)
for (i in 1:length(Vlabels)) {
  tmp = cor.test(Dtmp$pcu,Dtmp[,i])
  Rcorr[i,1] = tmp$estimate
  Rcorr[i,2] = tmp$conf.int[1]
  Rcorr[i,3] = tmp$conf.int[2]
  tmpBF = correlationBF(y = Dtmp$pcu, x = Dtmp[,i], rscale = 'ultrawide', progress = FALSE)
  # ultrawide = beta is 1 (default in JASP)
  Rcorr[i,4] = 1/extractBF(tmpBF, onlybf = TRUE)
  tmpsamples = posterior(tmpBF, iterations = 10000, progress = FALSE)
  tmppost = summary(tmpsamples)
  Rcorr[i,5] = tmppost$quantiles[1,1]
  Rcorr[i,6] = tmppost$quantiles[1,5]
  Rcorr[i,7] = length(Dtmp[,i])
}
Rcorr = round(Rcorr, digits = 3)
Rcorr = data.frame(Vlabels, Rcorr)
colnames(Rcorr) <- c('Variable', 'Correlation', 'LL', 'UL', 'BF01', 'Cred_LL', 'Cred_UL', 'N')
#write.table(Rcorr, file = paste0("table_WMC_corr_",nowsample,"_subjects.tsv"), row.names = F, dec = ".", sep = "\t")
# save variables for JASP
Dtmp3 = cbind(D$pcu,Dtmp)
#write.table(Dtmp3, file = paste0("data_WMC_EEG_",nowsample,"_subjects.tsv"), row.names = F, dec = ".", sep = "\t")

Rcorr %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Correlation	LL	UL	BF01	Cred_LL	Cred_UL	N
AmpSNRlo	0.069	-0.257	0.380	4.566	-0.248	0.363	38
AmpSlo	-0.037	-0.353	0.286	4.835	-0.340	0.282	38
AmpSmNlo	-0.012	-0.330	0.309	4.939	-0.317	0.297	38
ItcSNRlo	-0.014	-0.332	0.307	4.935	-0.323	0.302	38
ItcSlo	0.017	-0.304	0.335	4.926	-0.293	0.321	38
ItcSmNlo	-0.013	-0.332	0.308	4.936	-0.322	0.292	38
AmpSNRhi	-0.006	-0.325	0.314	4.947	-0.311	0.298	38
AmpShi	-0.022	-0.339	0.300	4.911	-0.321	0.295	38
AmpSmNhi	-0.002	-0.322	0.318	4.951	-0.316	0.303	38
ItcSNRhi	0.007	-0.313	0.326	4.947	-0.302	0.311	38
ItcShi	0.007	-0.314	0.326	4.947	-0.309	0.322	38
ItcSmNhi	-0.017	-0.334	0.305	4.928	-0.319	0.297	38
AmpSNRloMhi	-0.008	-0.327	0.313	4.946	-0.314	0.301	38
AmpSloMhi	-0.028	-0.345	0.294	4.883	-0.336	0.277	38
AmpSmNloMhi	-0.028	-0.345	0.294	4.884	-0.334	0.277	38
ItcSNRloMhi	-0.054	-0.368	0.270	4.707	-0.357	0.255	38
ItcSloMhi	0.017	-0.304	0.335	4.926	-0.292	0.318	38
ItcSmNloMhi	0.008	-0.313	0.326	4.946	-0.307	0.314	38

scatterplots

Scatterplots of working memory capacity (wmc.pcu) with EEG variables.

Dtmp = subset(Dtmp, select = c(AmpSloMhi, AmpSmNloMhi, ItcSloMhi, ItcSmNloMhi, pcu))
pairs(Dtmp)

effect of block

Explore factorial analyses (load x block).

by measure

Dtmp1a = subset(D, select=c(fp, 
                            AmpSNRlo1, AmpSNRlo2, AmpSNRlo3, AmpSNRlo4,
                            AmpSNRhi1, AmpSNRhi2, AmpSNRhi3, AmpSNRhi4))
Dtmp1b = subset(D, select=c(fp, 
                            AmpSlo1, AmpSlo2, AmpSlo3, AmpSlo4,
                            AmpShi1, AmpShi2, AmpShi3, AmpShi4))
Dtmp1c = subset(D, select=c(fp, 
                            AmpSmNlo1, AmpSmNlo2, AmpSmNlo3, AmpSmNlo4,
                            AmpSmNhi1, AmpSmNhi2, AmpSmNhi3, AmpSmNhi4))
Dtmp1d = subset(D, select=c(fp, 
                            ItcSNRlo1, ItcSNRlo2, ItcSNRlo3, ItcSNRlo4,
                            ItcSNRhi1, ItcSNRhi2, ItcSNRhi3, ItcSNRhi4))
Dtmp1e = subset(D, select=c(fp, 
                            ItcSlo1, ItcSlo2, ItcSlo3, ItcSlo4,
                            ItcShi1, ItcShi2, ItcShi3, ItcShi4))
Dtmp1f = subset(D, select=c(fp, 
                            ItcSmNlo1, ItcSmNlo2, ItcSmNlo3, ItcSmNlo4,
                            ItcSmNhi1, ItcSmNhi2, ItcSmNhi3, ItcSmNhi4))
Dtmp1g = subset(D, select=c(fp, 
                            vP3lo1, vP3lo2, vP3lo3, vP3lo4,
                            vP3hi1, vP3hi2, vP3hi3, vP3hi4))
Dtmp1 = list(Dtmp1a, Dtmp1b, Dtmp1c, Dtmp1d, Dtmp1e, Dtmp1f, Dtmp1g)
Dvars = c('Amplitude_SNR','Amplitude_S','Amplitude_SmN',
          'Intertrial_Coherence_SNR','Intertrial_Coherence_S','Intertrial_Coherence_SmN',
          'Visual_P3')
Fres = (matrix(ncol = 5, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','Overall_Interaction_p','Linear_Interaction_p','BF01_Interaction', 'BF01_error%')

amplitude SNR

Fres = computeANOVA(1, Dtmp1[[1]])

## =====================================================================
##                         Amplitude_SNR
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N     Mean    CI_LL    CI_UL
##   low     1 38 5.541296 4.617080 6.465512
##   low     2 38 5.589231 4.572741 6.605721
##   low     3 38 6.071998 5.055335 7.088661
##   low     4 38 5.892702 4.880950 6.904454
##  high     1 38 5.103951 4.149983 6.057919
##  high     2 38 5.158051 4.317833 5.998269
##  high     3 38 6.056604 4.942529 7.170679
##  high     4 38 5.574338 4.600451 6.548225
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 9613.7      1  2010.76     37 176.9013 1.139e-15 ***
## load           6.9      1   161.13     37   1.5767   0.21711    
## block         27.3      3   272.91    111   3.7028   0.01388 *  
## load:block     2.2      3   209.46    111   0.3941   0.75752    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.67966 0.01702
## load:block        0.84891 0.32113
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])  
## block      0.81143    0.02116 *
## load:block 0.91475    0.73983  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps Pr(>F[HF])
## block      0.8727594 0.01843815
## load:block 0.9952739 0.75658201
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                   Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 -0.12218  1 0.0063156  0.23516      1     37 0.6306
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 21.94085 ±5.54%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

amplitude S

Fres = computeANOVA(2, Dtmp1[[2]])

## =====================================================================
##                         Amplitude_S
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.2376145 0.2051571 0.2700719
##   low     2 38 0.2318844 0.1990759 0.2646930
##   low     3 38 0.2451609 0.2119793 0.2783424
##   low     4 38 0.2394622 0.2057286 0.2731957
##  high     1 38 0.2352709 0.2035864 0.2669555
##  high     2 38 0.2379926 0.2033864 0.2725987
##  high     3 38 0.2425984 0.2115264 0.2736705
##  high     4 38 0.2290730 0.1980479 0.2600980
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value Pr(>F)    
## (Intercept) 17.1305      1  2.49600     37 253.9372 <2e-16 ***
## load         0.0004      1  0.07640     37   0.1942 0.6620    
## block        0.0045      3  0.18226    111   0.9068 0.4403    
## load:block   0.0026      3  0.15579    111   0.6146 0.6069    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.89258 0.54107
## load:block        0.73108 0.04784
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])
## block      0.92885     0.4346
## load:block 0.81845     0.5755
## 
##               HF eps Pr(>F[HF])
## block      1.0121588  0.4403042
## load:block 0.8810143  0.5869969
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0051873  1  0.012467  0.46708      1     37 0.4986
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 15.21324 ±3.6%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

amplitude SmN

Fres = computeANOVA(3, Dtmp1[[3]])

## =====================================================================
##                         Amplitude_SmN
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.1917437 0.1586966 0.2247908
##   low     2 38 0.1860308 0.1521809 0.2198808
##   low     3 38 0.2006720 0.1679731 0.2333709
##   low     4 38 0.1955871 0.1621908 0.2289835
##  high     1 38 0.1834935 0.1502303 0.2167568
##  high     2 38 0.1884028 0.1530494 0.2237563
##  high     3 38 0.1968174 0.1651981 0.2284368
##  high     4 38 0.1831053 0.1515181 0.2146924
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value   Pr(>F)    
## (Intercept) 11.0591      1  2.53639     37 161.3260 4.65e-15 ***
## load         0.0023      1  0.09559     37   0.9073   0.3470    
## block        0.0067      3  0.20255    111   1.2241   0.3044    
## load:block   0.0023      3  0.17110    111   0.4970   0.6851    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic  p-value
## block             0.81746 0.206349
## load:block        0.73577 0.052253
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])
## block      0.87643     0.3037
## load:block 0.82101     0.6487
## 
##               HF eps Pr(>F[HF])
## block      0.9495826   0.304238
## load:block 0.8840288   0.662326
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0029918  1 0.0038618  0.14344      1     37  0.707
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 17.70882 ±2.06%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC SNR

Fres = computeANOVA(4, Dtmp1[[4]])

## =====================================================================
##                         Intertrial_Coherence_SNR
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N     Mean    CI_LL    CI_UL
##   low     1 38 4.840972 4.168869 5.513076
##   low     2 38 4.743656 4.082517 5.404795
##   low     3 38 5.297175 4.669979 5.924371
##   low     4 38 5.142990 4.445107 5.840873
##  high     1 38 4.476288 3.750315 5.202261
##  high     2 38 4.578926 3.953844 5.204008
##  high     3 38 4.977394 4.263140 5.691649
##  high     4 38 4.942340 4.185994 5.698686
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value  Pr(>F)    
## (Intercept) 7224.7      1   922.52     37 289.7637 < 2e-16 ***
## load           5.2      1    81.24     37   2.3844 0.13106    
## block         14.4      3   159.22    111   3.3450 0.02179 *  
## load:block     0.5      3   127.88    111   0.1490 0.93012    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.81367 0.19497
## load:block        0.87697 0.45507
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])  
## block      0.87221    0.02765 *
## load:block 0.92386    0.91872  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps Pr(>F[HF])
## block      0.9445633 0.02415654
## load:block 1.0061778 0.93012482
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 -0.053293  1 0.0018916 0.070121      1     37 0.7926
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 26.05604 ±2.99%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC S

Fres = computeANOVA(5, Dtmp1[[5]])

## =====================================================================
##                         Intertrial_Coherence_S
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.3820340 0.3307349 0.4333331
##   low     2 38 0.3810665 0.3259302 0.4362029
##   low     3 38 0.4128330 0.3641120 0.4615541
##   low     4 38 0.4116266 0.3616756 0.4615775
##  high     1 38 0.3623725 0.3099345 0.4148104
##  high     2 38 0.3752397 0.3263081 0.4241714
##  high     3 38 0.3940858 0.3412093 0.4469622
##  high     4 38 0.3859725 0.3307252 0.4412199
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value  Pr(>F)    
## (Intercept) 45.802      1   5.9490     37 284.8665 < 2e-16 ***
## load         0.023      1   0.3010     37   2.8519 0.09968 .  
## block        0.053      3   0.6973    111   2.8311 0.04167 *  
## load:block   0.004      3   0.4271    111   0.3440 0.79354    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic  p-value
## block             0.76061 0.081993
## load:block        0.78959 0.133782
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])  
## block      0.83704     0.0519 .
## load:block 0.88335     0.7687  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps Pr(>F[HF])
## block      0.9029197 0.04748206
## load:block 0.9578076 0.78497562
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0048854  1 0.0049258  0.18316      1     37 0.6712
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 23.35404 ±4.67%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC SmN

Fres = computeANOVA(6, Dtmp1[[6]])

## =====================================================================
##                         Intertrial_Coherence_SmN
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.3020561 0.2508146 0.3532976
##   low     2 38 0.3003659 0.2457838 0.3549479
##   low     3 38 0.3342563 0.2862179 0.3822947
##   low     4 38 0.3298787 0.2794422 0.3803153
##  high     1 38 0.2786943 0.2252802 0.3321085
##  high     2 38 0.2915337 0.2426769 0.3403904
##  high     3 38 0.3131948 0.2601659 0.3662236
##  high     4 38 0.3058960 0.2497522 0.3620399
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 28.6488      1   5.9435     37 178.3466 1.005e-15 ***
## load         0.0283      1   0.3145     37   3.3337   0.07595 .  
## block        0.0606      3   0.6994    111   3.2034   0.02605 *  
## load:block   0.0029      3   0.4607    111   0.2306   0.87493    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic  p-value
## block             0.74782 0.065244
## load:block        0.79745 0.151746
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])  
## block      0.82887    0.03497 *
## load:block 0.89076    0.85366  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps Pr(>F[HF])
## block      0.8932852 0.03129378
## load:block 0.9666266 0.86884247
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df  test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0022281  1 0.00094623 0.035044      1     37 0.8525
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 23.18317 ±2.18%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

visual P3

Fres = computeANOVA(7, Dtmp1[[7]])

## =====================================================================
##                         Visual_P3
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 10.817428 8.4589435 13.175912
##   low     2 38  8.091289 5.8959825 10.286596
##   low     3 38  7.811407 5.7324605  9.890353
##   low     4 38  7.431825 5.7474716  9.116178
##  high     1 38  2.781675 1.1999090  4.363441
##  high     2 38  2.380418 0.5710812  4.189755
##  high     3 38  1.799989 0.0824797  3.517499
##  high     4 38  3.040522 1.4331476  4.647896
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df F value    Pr(>F)    
## (Intercept) 9260.7      1   7052.6     37 48.5845 3.106e-08 ***
## load        2770.2      1   1239.8     37 82.6742 5.742e-11 ***
## block        175.5      3    871.0    111  7.4541 0.0001361 ***
## load:block   129.4      3    717.0    111  6.6774 0.0003472 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.59174 0.00216
## load:block        0.87050 0.42168
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])    
## block      0.75248  0.0006505 ***
## load:block 0.91582  0.0005492 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps   Pr(>F[HF])
## block      0.8038469 0.0004694969
## load:block 0.9965570 0.0003537538
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                  Value Df test stat approx F num Df den Df   Pr(>F)   
## load1 : block1 -1.6812  1   0.24397    11.94      1     37 0.001395 **
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 0.2616009 ±1.57%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

summary

Summary of interaction analyses:

Overall_interaction has 3 dfs.
Linear_interaction has 1 df and captures the interaction of load with the linear trend over blocks.
BF01_interaction is the BF for a model without the overall interaction versus a model with the overall interaction (prior: Cauchy r scale = 0.5).
BF01_error% estimates the error of the BF in percent. For example, if BF = 10 and the error is 10%, then the BF is estimated to vary between 9 and 11.

write.table(Fres, file = file.path(dir_fig,
                                   paste0("table_ANOVA_by_block_interaction_p_",nowsample,"_subjects.tsv")),
                                   row.names = F, dec = ".", sep = "\t")
Fres[,2:5] = round(Fres[,2:5], digits = 3)
Fres[Fres < 0.001] = 0.001
Fres %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Overall_Interaction_p	Linear_Interaction_p	BF01_Interaction	BF01_error%
Amplitude_SNR	0.740	0.631	21.941	5.537
Amplitude_S	0.575	0.499	15.213	3.599
Amplitude_SmN	0.649	0.707	17.709	2.055
Intertrial_Coherence_SNR	0.919	0.793	26.056	2.991
Intertrial_Coherence_S	0.769	0.671	23.354	4.668
Intertrial_Coherence_SmN	0.854	0.853	23.183	2.180
Visual_P3	0.001	0.001	0.262	1.567

reliability

Reliability of EEG variables over blocks. Results show mean correlations between the four blocks.
Note that raw scores (i.e., low and high load) have decent reliability whereas the difference scores show low reliability. This is not necessarily a problem: If all subjects change similarly, any variability in the difference scores would reflect mainly measurement noise, and the difference scores would show a low correlations over blocks (i.e., low reliability). Because of their low reliability, the difference scores are a poor measure if the goal is to distinguish the order among individual subjects.

Cres = (matrix(ncol = 4, nrow = 4))
Cres = as.data.frame(Cres)
colnames(Cres) = c('Variable','low','high', 'loMhi')
Vlabels = c("ampSNR", "ampS", "ampSmN", "itcSNR", "itcS", "itcSmN")
for (i in 1:length(Vlabels)) {
  Cres[i,1] = Vlabels[i]
  Dtmp2 = Dtmp1[[i]]
  Dtmp2$fp = NULL
  # low  
  tmp = cor(Dtmp2[,1:4])
  diag(tmp) = NA
  tmp = FisherZ(tmp)
  Cres[i,2] = FisherZInv(mean(tmp, na.rm = TRUE))
  # high
  tmp = cor(Dtmp2[,5:8])
  diag(tmp) = NA
  tmp = FisherZ(tmp)
  Cres[i,3] = FisherZInv(mean(tmp, na.rm = TRUE))
  # low vs high
  tmp = Dtmp2[,1:4] - Dtmp2[,5:8]
  tmp = cor(tmp)
  diag(tmp) = NA
  tmp = FisherZ(tmp)
  Cres[i,4] = FisherZInv(mean(tmp, na.rm = TRUE))
}
#write.table(Cres, file = paste0("table_Mean_Corr_Blocks_",nowsample,"_subjects.tsv"), row.names = F, dec = ".", sep = "\t")
Cres %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	low	high	loMhi
ampSNR	0.748	0.792	0.247
ampS	0.836	0.865	0.114
ampSmN	0.820	0.860	0.157
itcSNR	0.712	0.710	0.193
itcS	0.808	0.802	0.225
itcSmN	0.800	0.800	0.215

effect of minute

Analysis of EEG data in 1-min blocks.
The data in each 3-min block were analyzed by minute (this was done during preprocessing in MNE-python). For each minute, mean amp (and itc) were computed and then averaged across the four blocks of load.
Note that the R output uses the term “block,” but this is meant to refer to “minute.”

by measure

# The data in each 3-min block are analyzed by minute. For each minute, the amp and itc are computed and then averaged across the four blocks of load.  
Dtmp1a = subset(D, select=c(fp, 
                            m1loampSNR, m2loampSNR, m3loampSNR,
                            m1hiampSNR, m2hiampSNR, m3hiampSNR))
Dtmp1b = subset(D, select=c(fp, 
                            m1loampS, m2loampS, m3loampS,
                            m1hiampS, m2hiampS, m3hiampS))
Dtmp1c = subset(D, select=c(fp, 
                            m1loampSmN, m2loampSmN, m3loampSmN,
                            m1hiampSmN, m2hiampSmN, m3hiampSmN))
Dtmp1d = subset(D, select=c(fp, 
                            m1loitcSNR, m2loitcSNR, m3loitcSNR,
                            m1hiitcSNR, m2hiitcSNR, m3hiitcSNR))
Dtmp1e = subset(D, select=c(fp, 
                            m1loitcS, m2loitcS, m3loitcS,
                            m1hiitcS, m2hiitcS, m3hiitcS))
Dtmp1f = subset(D, select=c(fp, 
                            m1loitcSmN, m2loitcSmN, m3loitcSmN,
                            m1hiitcSmN, m2hiitcSmN, m3hiitcSmN))
Dtmp1 = list(Dtmp1a, Dtmp1b, Dtmp1c, Dtmp1d, Dtmp1e, Dtmp1f)
# rename variables to be able to call computeANOVA
for (i in 1:6){
  tmp = Dtmp1[[i]]
  colnames(tmp) <- c('fp','lo1','lo2','lo3','hi1','hi2','hi3')
  Dtmp1[[i]] = tmp
}
Dvars = c('Amplitude_SNR','Amplitude_S','Amplitude_SmN', 
          'Intertrial_Coherence_SNR', 'Intertrial_Coherence_S','Intertrial_Coherence_SmN')
Fres = (matrix(ncol = 5, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','Overall_Interaction_p','Linear_Interaction_p','BF01_Interaction', 'BF01_error%')

amplitude SNR

Fres = computeANOVA(1, Dtmp1[[1]], "minute")

## =====================================================================
##                         Amplitude_SNR
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N     Mean    CI_LL    CI_UL
##   low     1 38 4.333495 3.650796 5.016195
##   low     2 38 3.936829 3.330898 4.542760
##   low     3 38 3.786145 3.251514 4.320776
##  high     1 38 4.027695 3.420574 4.634815
##  high     2 38 3.698736 3.099608 4.297864
##  high     3 38 3.560498 2.973739 4.147258
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 3451.1      1   676.55     37 188.7383 4.185e-16 ***
## load           3.8      1    28.08     37   4.9423    0.0324 *  
## block         10.4      2    30.78     74  12.4789 2.138e-05 ***
## load:block     0.1      2    14.90     74   0.1755    0.8394    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.94649 0.37158
## load:block        0.93631 0.30589
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])    
## block      0.94920  3.187e-05 ***
## load:block 0.94012     0.8264    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##               HF eps   Pr(>F[HF])
## block      0.9990927 2.152936e-05
## load:block 0.9887518 8.370230e-01
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 -0.040077  1 0.0066505  0.24771      1     37 0.6216
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 11.64508 ±1.44%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

amplitude S

Fres = computeANOVA(2, Dtmp1[[2]], "minute")

## =====================================================================
##                         Amplitude_S
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.2888944 0.2528880 0.3249008
##   low     2 38 0.2842369 0.2467874 0.3216865
##   low     3 38 0.2754078 0.2425218 0.3082937
##  high     1 38 0.2865101 0.2492056 0.3238146
##  high     2 38 0.2810947 0.2445676 0.3176218
##  high     3 38 0.2683740 0.2351294 0.3016186
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value  Pr(>F)    
## (Intercept) 17.9715      1  2.38762     37 278.4965 < 2e-16 ***
## load         0.0010      1  0.05907     37   0.6258 0.43394    
## block        0.0099      2  0.10300     74   3.5623 0.03334 *  
## load:block   0.0002      2  0.05701     74   0.1535 0.85800    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.95904 0.47106
## load:block        0.97163 0.59567
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])  
## block      0.96065    0.03524 *
## load:block 0.97241    0.85239  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##              HF eps Pr(>F[HF])
## block      1.012148 0.03333713
## load:block 1.025573 0.85799941
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0023247  1 0.0078459  0.29259      1     37 0.5918
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 11.14701 ±1.43%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

amplitude SmN

Fres = computeANOVA(3, Dtmp1[[3]], "minute")

## =====================================================================
##                         Amplitude_SmN
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.2159262 0.1789746 0.2528777
##   low     2 38 0.2042131 0.1671492 0.2412769
##   low     3 38 0.1957877 0.1626295 0.2289459
##  high     1 38 0.2081259 0.1702517 0.2460002
##  high     2 38 0.1960791 0.1588174 0.2333408
##  high     3 38 0.1827327 0.1477186 0.2177467
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 9.1636      1  2.46984     37 137.2775 5.159e-14 ***
## load        0.0053      1  0.07005     37   2.8111  0.102045    
## block       0.0197      2  0.10675     74   6.8307  0.001895 ** 
## load:block  0.0003      2  0.05460     74   0.2229  0.800707    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.92600 0.25060
## load:block        0.98658 0.78408
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])   
## block      0.93110   0.002463 **
## load:block 0.98676   0.797836   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##              HF eps  Pr(>F[HF])
## block      0.978482 0.002056694
## load:block 1.041974 0.800707429
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0026274  1 0.0099608  0.37226      1     37 0.5455
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 10.73122 ±3.73%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC SNR

Fres = computeANOVA(4, Dtmp1[[4]], "minute")

## =====================================================================
##                         Intertrial_Coherence_SNR
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N     Mean    CI_LL    CI_UL
##   low     1 38 3.603809 3.191972 4.015647
##   low     2 38 3.388385 2.990486 3.786285
##   low     3 38 3.195458 2.830853 3.560063
##  high     1 38 3.372475 2.957167 3.787782
##  high     2 38 3.129813 2.740749 3.518877
##  high     3 38 3.058135 2.661785 3.454485
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 2469.91      1  287.708     37 317.6379 < 2.2e-16 ***
## load           2.49      1   10.815     37   8.5241  0.005935 ** 
## block          5.08      2   15.209     74  12.3587 2.339e-05 ***
## load:block     0.15      2    8.812     74   0.6457  0.527232    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.98003 0.69551
## load:block        0.97860 0.67746
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])    
## block      0.98042  2.723e-05 ***
## load:block 0.97905     0.5241    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##              HF eps   Pr(>F[HF])
## block      1.034727 2.338807e-05
## load:block 1.033157 5.272317e-01
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 -0.047006  1  0.018158  0.68425      1     37 0.4134
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 8.823038 ±2.03%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC S

Fres = computeANOVA(5, Dtmp1[[5]], "minute")

## =====================================================================
##                         Intertrial_Coherence_S
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.4927185 0.4357066 0.5497304
##   low     2 38 0.4729413 0.4190717 0.5268109
##   low     3 38 0.4471333 0.3968485 0.4974181
##  high     1 38 0.4698906 0.4109699 0.5288113
##  high     2 38 0.4435461 0.3896617 0.4974305
##  high     3 38 0.4199218 0.3667954 0.4730482
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 47.762      1   5.5738     37 317.0541 < 2.2e-16 ***
## load         0.040      1   0.1675     37   8.8251  0.005194 ** 
## block        0.087      2   0.2332     74  13.7686  8.25e-06 ***
## load:block   0.000      2   0.1483     74   0.1060  0.899543    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.99832 0.97024
## load:block        0.94134 0.33688
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])    
## block      0.99833  8.371e-06 ***
## load:block 0.94459     0.8894    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##              HF eps   Pr(>F[HF])
## block      1.055224 8.249580e-06
## load:block 0.993842 8.984641e-01
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                    Value Df test stat approx F num Df den Df Pr(>F)
## load1 : block1 0.0021918  1 0.0021791 0.080804      1     37 0.7778
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 10.24828 ±8.64%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

ITC SmN

Fres = computeANOVA(6, Dtmp1[[6]], "minute")

## =====================================================================
##                         Intertrial_Coherence_SmN
## =====================================================================
## 
## 
## Descriptives
## ============
##  load block  N      Mean     CI_LL     CI_UL
##   low     1 38 0.3545864 0.2979165 0.4112563
##   low     2 38 0.3303767 0.2759573 0.3847961
##   low     3 38 0.3045146 0.2541961 0.3548331
##  high     1 38 0.3287244 0.2698754 0.3875735
##  high     2 38 0.2994724 0.2452751 0.3536697
##  high     3 38 0.2796476 0.2259642 0.3333311
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df  F value    Pr(>F)    
## (Intercept) 22.7989      1   5.6051     37 150.4992 1.322e-14 ***
## load         0.0422      1   0.1746     37   8.9424  0.004933 ** 
## block        0.0936      2   0.2350     74  14.7327 4.113e-06 ***
## load:block   0.0004      2   0.1454     74   0.1013  0.903788    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## Mauchly Tests for Sphericity
## 
##            Test statistic p-value
## block             0.99848 0.97300
## load:block        0.95463 0.43351
## 
## 
## Greenhouse-Geisser and Huynh-Feldt Corrections
##  for Departure from Sphericity
## 
##             GG eps Pr(>F[GG])    
## block      0.99848  4.172e-06 ***
## load:block 0.95660      0.896    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
##              HF eps   Pr(>F[HF])
## block      1.055404 4.112897e-06
## load:block 1.007520 9.037883e-01
## 
## 
## Linear interaction contrast
## ===========================
## Multivariate Test: Pillai test statistic
## P-value adjustment method: holm
##                      Value Df  test stat  approx F num Df den Df Pr(>F)
## load1 : block1 -0.00049752  1 0.00011648 0.0043101      1     37  0.948
## 
## 
## 
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + block + fp : 11.52614 ±3.12%
## 
## Against denominator:
##   eeg_dv ~ load + block + load:block + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

summary

Summary of interaction analyses:

Overall_interaction has 3 dfs.
Linear_interaction has 1 df and captures the interaction of load with the linear trend over blocks.
BF01_interaction is the BF for a model without the overall interaction versus a model with the overall interaction (prior: Cauchy r scale = 0.5).
BF01_error% estimates the error of the BF in percent. For example, if BF = 10 and the error is 10%, then the BF is estimated to vary between 9 and 11.

write.table(Fres, file = file.path(dir_fig,
 paste0("table_ANOVA_by_minute_interaction_p_",nowsample,"_subjects.tsv")),
                                  row.names = F, dec = ".", sep = "\t")
Fres[,2:5] = round(Fres[,2:5], digits = 3)
Fres[Fres < 0.001] = 0.001
Fres %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	Overall_Interaction_p	Linear_Interaction_p	BF01_Interaction	BF01_error%
Amplitude_SNR	0.826	0.622	11.645	1.443
Amplitude_S	0.852	0.592	11.147	1.425
Amplitude_SmN	0.798	0.546	10.731	3.735
Intertrial_Coherence_SNR	0.524	0.413	8.823	2.033
Intertrial_Coherence_S	0.889	0.778	10.248	8.636
Intertrial_Coherence_SmN	0.896	0.948	11.526	3.117

Dtmp = D
Dtmp$study = 'Study1'
Dtmp = subset(Dtmp, select=c(study, fp,
  AmpSmNlo, AmpSmNhi, ItcSmNlo, ItcSmNhi))
write.table(Dtmp, paste0("data_export_SmN_ASSR_study1_",nowsample,"_subjects.tsv"), sep="\t", row.names = FALSE)


Dtmp = D 
Dtmp$study = 'Study1'
Dtmp = subset(Dtmp, 
              select=c(fp, 
                      AmpSmNlo1, AmpSmNlo2, AmpSmNlo3, AmpSmNlo4,
                      AmpSmNhi1, AmpSmNhi2, AmpSmNhi3, AmpSmNhi4,
                      ItcSmNlo1, ItcSmNlo2, ItcSmNlo3, ItcSmNlo4,
                      ItcSmNhi1, ItcSmNhi2, ItcSmNhi3, ItcSmNhi4,
                      m1loampSmN, m2loampSmN, m3loampSmN,
                      m1hiampSmN, m2hiampSmN, m3hiampSmN,
                      m1loitcSmN, m2loitcSmN, m3loitcSmN,
                      m1hiitcSmN, m2hiitcSmN, m3hiitcSmN))
write.table(Dtmp, paste0("data_export_SmN_by_block_ASSR_study1_",nowsample,"_subjects.tsv"), sep="\t", row.names = FALSE)

time-frequency analysis at 2 Hz

The goal of this exploratory time-frequency analysis was to examine if the 40-Hz signal changed periodically with the onset of the visual stimuli, which were shown every 500 ms (i.e., at 2 Hz). To detect this period change, epochs were rather long (10 s), and each epoch began every 20th visual stimulus (see MNE-python script).

These epochs should contain a 40-Hz signal (even though the phase shifts between epochs).
If the onsets of the visual stimuli affect the 40-Hz signal periodically, then the 40-Hz signal should change at 2 Hz within an epoch.
Thus, within the 40-Hz signal, there should be a signal at 2 Hz.
The noise can be defined by surrounding frequencies (10 on each side excluding the 2 nearest neighbors).
Results show amplitude SmN and S for the 2-Hz signal.

A potential concern is that the 40-Hz response may be confounded by indirect visual effects on the same electrodes as used for ASSRs. That is, the electrodes that were used to record ASSRs may pick up unrelated ERP activity from visual onsets. However, because the response is recorded at 40 Hz whereas visual events occur at 2 Hz, an analysis of only the 40-Hz response should already remove confounding effects of the visual events at 2 Hz (because the frequency is much lower). Accordingly, any 2-Hz activity within the 40-Hz response suggests that the 40-Hz response is actually affected by the visual onsets.

Although results suggest some activity at 2 Hz, there were no differences among the load conditions.

figure of amplitude spectrum

The figure (N = 43) shows amplitude over time (top) and spectrum (bottom). The figure was generated in MNE-python (see script).

knitr::include_graphics("40 Hz amplitude over time and spectrum for 10s_ind.jpeg")

Amplitude and spectrum

Dtf = read.csv("data_audio_signal_noise_tf_amplitude.tsv", sep="\t", header=TRUE, dec=".")
Dvars = c('AmpSmN','AmpS')
Fres = (matrix(ncol = 4, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','P_load','BF01_load', 'BF01_error%')

# maybe exclude subjects? see above
if (nowsample == "subset") {
  excludefps = c(3, 20, 24, 35, 40)
  indx = Dtf$fp %in% excludefps
  Dtf = Dtf[!indx,]}
  #table(D$fp)

by measure

amplitude SmN

N = 38

Fres = computeANOVAtf(Dtf, 1)

## =====================================================================
##                         AmpSmN
## =====================================================================
## 
## 
## Descriptives
## ============
##  load  N     Mean     CI_LL    CI_UL
##   low 38 1.592818 0.7446397 2.440996
##  high 38 1.707334 0.8536819 2.560986
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##              Sum Sq num Df Error SS den Df F value   Pr(>F)    
## (Intercept) 206.929      1   446.05     37 17.1649 0.000191 ***
## load          0.249      1    49.89     37  0.1848 0.669796    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## BF of main effect
## =================
## Bayes factor analysis
## --------------
## [1] fp : 3.965507 ±0.59%
## 
## Against denominator:
##   eeg_dv ~ load + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

amplitude S

N = 38

Fres = computeANOVAtf(Dtf, 2)

## =====================================================================
##                         AmpS
## =====================================================================
## 
## 
## Descriptives
## ============
##  load  N     Mean    CI_LL    CI_UL
##   low 38 3.182054 2.292691 4.071417
##  high 38 3.208384 2.307903 4.108865
## 
## 
## Anova
## =====
## 
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
## 
##             Sum Sq num Df Error SS den Df F value    Pr(>F)    
## (Intercept) 775.92      1   500.01     37  57.416 4.891e-09 ***
## load          0.01      1    48.57     37   0.010    0.9207    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## 
## BF of main effect
## =================
## Bayes factor analysis
## --------------
## [1] fp : 4.204424 ±1.84%
## 
## Against denominator:
##   eeg_dv ~ load + fp 
## ---
## Bayes factor type: BFlinearModel, JZS

summary

Summary of 2-Hz tf analysis:

P_load is the p value (from the ANOVA) for the main effect of load with 3 dfs.
BF01_load is the BF for a model without the main effect versus a model with the main effect (prior: Cauchy r scale = 0.5).
BF01_error% estimates the error of the BF in percent. For example, if BF = 10 and the error is 10%, then the BF is estimated to vary between 9 and 11.

N = 38

Fres[,2:4] = round(Fres[,2:4], digits = 3)
Fres[Fres < 0.001] = 0.001
Fres %>%
  kable(digits = 3) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), 
                full_width = F, position = 'left')

Variable	P_load	BF01_load	BF01_error%
AmpSmN	0.670	3.966	0.589
AmpS	0.921	4.204	1.842

ASSR Study 1 (2020)

Stefan Wiens & Malina Szychowska

2020-08-24

define sample

quick data check

hit rates

tables

behavior

EEG

combined

Bayes factors for load

switch from SNR to SmN

explanation

SNR is noisy

SmN is better

plot amp and ITC

correlations with behavior

table

scatterplots

fishing

correlations with wm capacity

table

scatterplots

effect of block

by measure

amplitude SNR

amplitude S

amplitude SmN

ITC SNR

ITC S

ITC SmN

visual P3

summary

reliability

effect of minute

by measure

amplitude SNR

amplitude S

amplitude SmN

ITC SNR

ITC S

ITC SmN

summary

time-frequency analysis at 2 Hz

figure of amplitude spectrum

by measure

amplitude SmN

amplitude S

summary