ASSR2
Stefan Wiens & Malina Szychowska
2020-09-28
Define sample
Note: The analyses use the sample of all subjects.
For all subjects, 8 pilot subjects were included. These were tested before formal preregistration.
For subset subjects, 8 pilot subjects were excluded.
tmpaEEG=read.csv(file.path(dir_res,"data_audio_signal_noise_N_33.tsv"), sep="\t",header=TRUE, dec=".")
#head(tmpaEEG)# Analysis of visual P3
tmpvEEG=read.csv(file.path(dir_res,"data_visualP3_N_33.tsv"), sep="\t",header=TRUE, dec=".")
#head(tmpwmc)tmpbeh=read.csv(file.path(dir_res,"data_logfiles.tsv"), sep="\t",header=TRUE, dec=".")
#head(tmpbeh)Quick analyses
Sample size: N = 33.
Sanity check for behavioral performance (across low and high load, and across frequencies):
Min number of hits (could be 0) : 41
Max number of hits (could be 96) : 96
Min number of false alarms (could be 0) : 0
Max number of false alarms (could be 302) : 68
Note that hits are correct responses between 200 and 1000 ms after target onset (defined in ASSR2_process_beh.R)
False alarms are responses at other times: 199*2 (number of nontarget trials times 2 blocks) - 96 (trial after a target) = 302.
D$HRlo20Hzcorr = (D$nHRlo20Hz + 0.5)/(96+1)
D$HRlo40Hzcorr = (D$nHRlo40Hz + 0.5)/(96+1)
D$HRlo80Hzcorr = (D$nHRlo80Hz + 0.5)/(96+1)
D$HRhi20Hzcorr = (D$nHRhi20Hz + 0.5)/(96+1)
D$HRhi40Hzcorr = (D$nHRhi40Hz + 0.5)/(96+1)
D$HRhi80Hzcorr = (D$nHRhi80Hz + 0.5)/(96+1)
# 48*2 = 96 targets
D$FAlo20Hzcorr = (D$nFAlo20Hz + 0.5)/(398+1-96)
D$FAlo40Hzcorr = (D$nFAlo40Hz + 0.5)/(398+1-96)
D$FAlo80Hzcorr = (D$nFAlo80Hz + 0.5)/(398+1-96)
D$FAhi20Hzcorr = (D$nFAhi20Hz + 0.5)/(398+1-96)
D$FAhi40Hzcorr = (D$nFAhi40Hz + 0.5)/(398+1-96)
D$FAhi80Hzcorr = (D$nFAhi80Hz + 0.5)/(398+1-96)
# 199*2 = 398 nontargets
# do not count the next trial after a target (because this is considered to be a valid response)
# Dprime
D$dprlo20Hz = qnorm(D$HRlo20Hzcorr) - qnorm(D$FAlo20Hzcorr)
D$dprlo40Hz = qnorm(D$HRlo40Hzcorr) - qnorm(D$FAlo40Hzcorr)
D$dprlo80Hz = qnorm(D$HRlo80Hzcorr) - qnorm(D$FAlo80Hzcorr)
D$dprhi20Hz = qnorm(D$HRhi20Hzcorr) - qnorm(D$FAhi20Hzcorr)
D$dprhi40Hz = qnorm(D$HRhi40Hzcorr) - qnorm(D$FAhi40Hzcorr)
D$dprhi80Hz = qnorm(D$HRhi80Hzcorr) - qnorm(D$FAhi80Hzcorr)
# do not compute criterion
#tmp$criterion = - (1/2) * (qnorm(tmp$aware_crit_corrected) + qnorm(tmp$aware_catch_corrected))
D$'dprlo-hi20Hz' = D$dprlo20Hz - D$dprhi20Hz
D$'dprlo-hi40Hz' = D$dprlo40Hz - D$dprhi40Hz
D$'dprlo-hi80Hz' = D$dprlo80Hz - D$dprhi80Hz
D$'meanRTmslo-hi20Hz' = D$meanRTmslo20Hz - D$meanRTmshi20Hz
D$'meanRTmslo-hi40Hz' = D$meanRTmslo40Hz - D$meanRTmshi40Hz
D$'meanRTmslo-hi80Hz' = D$meanRTmslo80Hz - D$meanRTmshi80Hz# the numbers 1 to 2 refer to different blocks
# Amp S-N raw
D$AmpSmNno20Hz = rowMeans(cbind(D$AmpSmNno20_1,D$AmpSmNno20_2))
D$AmpSmNno40Hz = rowMeans(cbind(D$AmpSmNno40_1,D$AmpSmNno40_2))
D$AmpSmNno80Hz = rowMeans(cbind(D$AmpSmNno80_1,D$AmpSmNno80_2))
D$AmpSmNlo20Hz = rowMeans(cbind(D$AmpSmNlo20_1,D$AmpSmNlo20_2))
D$AmpSmNlo40Hz = rowMeans(cbind(D$AmpSmNlo40_1,D$AmpSmNlo40_2))
D$AmpSmNlo80Hz = rowMeans(cbind(D$AmpSmNlo80_1,D$AmpSmNlo80_2))
D$AmpSmNhi20Hz = rowMeans(cbind(D$AmpSmNhi20_1,D$AmpSmNhi20_2))
D$AmpSmNhi40Hz = rowMeans(cbind(D$AmpSmNhi40_1,D$AmpSmNhi40_2))
D$AmpSmNhi80Hz = rowMeans(cbind(D$AmpSmNhi80_1,D$AmpSmNhi80_2))
# Amp S raw
D$AmpSno20Hz = rowMeans(cbind(D$AmpSno20_1,D$AmpSno20_2))
D$AmpSno40Hz = rowMeans(cbind(D$AmpSno40_1,D$AmpSno40_2))
D$AmpSno80Hz = rowMeans(cbind(D$AmpSno80_1,D$AmpSno80_2))
D$AmpSlo20Hz = rowMeans(cbind(D$AmpSlo20_1,D$AmpSlo20_2))
D$AmpSlo40Hz = rowMeans(cbind(D$AmpSlo40_1,D$AmpSlo40_2))
D$AmpSlo80Hz = rowMeans(cbind(D$AmpSlo80_1,D$AmpSlo80_2))
D$AmpShi20Hz = rowMeans(cbind(D$AmpShi20_1,D$AmpShi20_2))
D$AmpShi40Hz = rowMeans(cbind(D$AmpShi40_1,D$AmpShi40_2))
D$AmpShi80Hz = rowMeans(cbind(D$AmpShi80_1,D$AmpShi80_2))
# Amp N raw
D$AmpNno20Hz = rowMeans(cbind(D$AmpNno20_1,D$AmpNno20_2))
D$AmpNno40Hz = rowMeans(cbind(D$AmpNno40_1,D$AmpNno40_2))
D$AmpNno80Hz = rowMeans(cbind(D$AmpNno80_1,D$AmpNno80_2))
D$AmpNlo20Hz = rowMeans(cbind(D$AmpNlo20_1,D$AmpNlo20_2))
D$AmpNlo40Hz = rowMeans(cbind(D$AmpNlo40_1,D$AmpNlo40_2))
D$AmpNlo80Hz = rowMeans(cbind(D$AmpNlo80_1,D$AmpNlo80_2))
D$AmpNhi20Hz = rowMeans(cbind(D$AmpNhi20_1,D$AmpNhi20_2))
D$AmpNhi40Hz = rowMeans(cbind(D$AmpNhi40_1,D$AmpNhi40_2))
D$AmpNhi80Hz = rowMeans(cbind(D$AmpNhi80_1,D$AmpNhi80_2))
# Amp SNR raw
D$AmpSNRno20Hz = rowMeans(cbind(D$AmpSNRno20_1,D$AmpSNRno20_2))
D$AmpSNRno40Hz = rowMeans(cbind(D$AmpSNRno40_1,D$AmpSNRno40_2))
D$AmpSNRno80Hz = rowMeans(cbind(D$AmpSNRno80_1,D$AmpSNRno80_2))
D$AmpSNRlo20Hz = rowMeans(cbind(D$AmpSNRlo20_1,D$AmpSNRlo20_2))
D$AmpSNRlo40Hz = rowMeans(cbind(D$AmpSNRlo40_1,D$AmpSNRlo40_2))
D$AmpSNRlo80Hz = rowMeans(cbind(D$AmpSNRlo80_1,D$AmpSNRlo80_2))
D$AmpSNRhi20Hz = rowMeans(cbind(D$AmpSNRhi20_1,D$AmpSNRhi20_2))
D$AmpSNRhi40Hz = rowMeans(cbind(D$AmpSNRhi40_1,D$AmpSNRhi40_2))
D$AmpSNRhi80Hz = rowMeans(cbind(D$AmpSNRhi80_1,D$AmpSNRhi80_2))# the numbers 1 to 2 refer to different blocks
# ITC S-N raw
D$ItcSmNno20Hz = rowMeans(cbind(D$ItcSmNno20_1,D$ItcSmNno20_2))
D$ItcSmNno40Hz = rowMeans(cbind(D$ItcSmNno40_1,D$ItcSmNno40_2))
D$ItcSmNno80Hz = rowMeans(cbind(D$ItcSmNno80_1,D$ItcSmNno80_2))
D$ItcSmNlo20Hz = rowMeans(cbind(D$ItcSmNlo20_1,D$ItcSmNlo20_2))
D$ItcSmNlo40Hz = rowMeans(cbind(D$ItcSmNlo40_1,D$ItcSmNlo40_2))
D$ItcSmNlo80Hz = rowMeans(cbind(D$ItcSmNlo80_1,D$ItcSmNlo80_2))
D$ItcSmNhi20Hz = rowMeans(cbind(D$ItcSmNhi20_1,D$ItcSmNhi20_2))
D$ItcSmNhi40Hz = rowMeans(cbind(D$ItcSmNhi40_1,D$ItcSmNhi40_2))
D$ItcSmNhi80Hz = rowMeans(cbind(D$ItcSmNhi80_1,D$ItcSmNhi80_2))
# ITC S raw
D$ItcSno20Hz = rowMeans(cbind(D$ItcSno20_1,D$ItcSno20_2))
D$ItcSno40Hz = rowMeans(cbind(D$ItcSno40_1,D$ItcSno40_2))
D$ItcSno80Hz = rowMeans(cbind(D$ItcSno80_1,D$ItcSno80_2))
D$ItcSlo20Hz = rowMeans(cbind(D$ItcSlo20_1,D$ItcSlo20_2))
D$ItcSlo40Hz = rowMeans(cbind(D$ItcSlo40_1,D$ItcSlo40_2))
D$ItcSlo80Hz = rowMeans(cbind(D$ItcSlo80_1,D$ItcSlo80_2))
D$ItcShi20Hz = rowMeans(cbind(D$ItcShi20_1,D$ItcShi20_2))
D$ItcShi40Hz = rowMeans(cbind(D$ItcShi40_1,D$ItcShi40_2))
D$ItcShi80Hz = rowMeans(cbind(D$ItcShi80_1,D$ItcShi80_2))
# ITC N raw
D$ItcNno20Hz = rowMeans(cbind(D$ItcNno20_1,D$ItcNno20_2))
D$ItcNno40Hz = rowMeans(cbind(D$ItcNno40_1,D$ItcNno40_2))
D$ItcNno80Hz = rowMeans(cbind(D$ItcNno80_1,D$ItcNno80_2))
D$ItcNlo20Hz = rowMeans(cbind(D$ItcNlo20_1,D$ItcNlo20_2))
D$ItcNlo40Hz = rowMeans(cbind(D$ItcNlo40_1,D$ItcNlo40_2))
D$ItcNlo80Hz = rowMeans(cbind(D$ItcNlo80_1,D$ItcNlo80_2))
D$ItcNhi20Hz = rowMeans(cbind(D$ItcNhi20_1,D$ItcNhi20_2))
D$ItcNhi40Hz = rowMeans(cbind(D$ItcNhi40_1,D$ItcNhi40_2))
D$ItcNhi80Hz = rowMeans(cbind(D$ItcNhi80_1,D$ItcNhi80_2))
# ITC SNR raw
D$ItcSNRno20Hz = rowMeans(cbind(D$ItcSNRno20_1,D$ItcSNRno20_2))
D$ItcSNRno40Hz = rowMeans(cbind(D$ItcSNRno40_1,D$ItcSNRno40_2))
D$ItcSNRno80Hz = rowMeans(cbind(D$ItcSNRno80_1,D$ItcSNRno80_2))
D$ItcSNRlo20Hz = rowMeans(cbind(D$ItcSNRlo20_1,D$ItcSNRlo20_2))
D$ItcSNRlo40Hz = rowMeans(cbind(D$ItcSNRlo40_1,D$ItcSNRlo40_2))
D$ItcSNRlo80Hz = rowMeans(cbind(D$ItcSNRlo80_1,D$ItcSNRlo80_2))
D$ItcSNRhi20Hz = rowMeans(cbind(D$ItcSNRhi20_1,D$ItcSNRhi20_2))
D$ItcSNRhi40Hz = rowMeans(cbind(D$ItcSNRhi40_1,D$ItcSNRhi40_2))
D$ItcSNRhi80Hz = rowMeans(cbind(D$ItcSNRhi80_1,D$ItcSNRhi80_2))# visual P3 is difference between targets and nontargets.
# the numbers 1 to 4 refer to different blocks
D$vP3lo20 = rowMeans(cbind(D$vP3_lo_tar20_1 - D$vP3_lo_non20_1, D$vP3_lo_tar20_2 - D$vP3_lo_non20_2))
D$vP3lo40 = rowMeans(cbind(D$vP3_lo_tar40_1 - D$vP3_lo_non40_1, D$vP3_lo_tar40_2 - D$vP3_lo_non40_2))
D$vP3lo80 = rowMeans(cbind(D$vP3_lo_tar80_1 - D$vP3_lo_non80_1, D$vP3_lo_tar80_2 - D$vP3_lo_non80_2))
D$vP3hi20 = rowMeans(cbind(D$vP3_hi_tar20_1 - D$vP3_hi_non20_1, D$vP3_hi_tar20_2 - D$vP3_hi_non20_2))
D$vP3hi40 = rowMeans(cbind(D$vP3_hi_tar40_1 - D$vP3_hi_non40_1, D$vP3_hi_tar40_2 - D$vP3_hi_non40_2))
D$vP3hi80 = rowMeans(cbind(D$vP3_hi_tar80_1 - D$vP3_hi_non80_1, D$vP3_hi_tar80_2 - D$vP3_hi_non80_2))
D$'vP3lo-hi20' = D$vP3lo20 - D$vP3hi20
D$'vP3lo-hi40' = D$vP3lo40 - D$vP3hi40
D$'vP3lo-hi80' = D$vP3lo80 - D$vP3hi80Quick look at behavioral performance:
Simple hit rates (%) were as follows:
20-Hz AM:
Low load = 95.55
High load = 76.99
40-Hz AM:
Low load = 96.14
High load = 80.49
80-Hz AM:
Low load = 94.99
High load = 79.4
Manipulation checks of load
lo = low load
hi = high load
lo-hi = low minus high load
20, 40, 80 Hz = amplitude modulation frequency
LL and UL refer to the 95% confidence interval (from two-tailed t tests).
d prime and reaction time
dpr = d prime
meanRTms = RT to hits (ms)
Note: Because there was no task during no load, there were no targets and nontargets to examine d prime and RT.
table
Dtmp = subset(D, select=c('dprlo20Hz', 'dprhi20Hz', 'dprlo-hi20Hz',
'dprlo40Hz', 'dprhi40Hz', 'dprlo-hi40Hz',
'dprlo80Hz', 'dprhi80Hz', 'dprlo-hi80Hz',
'meanRTmslo20Hz', 'meanRTmshi20Hz', 'meanRTmslo-hi20Hz',
'meanRTmslo40Hz', 'meanRTmshi40Hz', 'meanRTmslo-hi40Hz',
'meanRTmslo80Hz', 'meanRTmshi80Hz', 'meanRTmslo-hi80Hz'))
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 |
|---|---|---|---|---|
| dprlo20Hz | 4.370 | 4.074 | 4.667 | 33 |
| dprhi20Hz | 2.483 | 2.220 | 2.745 | 33 |
| dprlo-hi20Hz | 1.888 | 1.582 | 2.193 | 33 |
| dprlo40Hz | 4.355 | 4.111 | 4.598 | 33 |
| dprhi40Hz | 2.569 | 2.345 | 2.794 | 33 |
| dprlo-hi40Hz | 1.785 | 1.560 | 2.011 | 33 |
| dprlo80Hz | 4.309 | 4.014 | 4.604 | 33 |
| dprhi80Hz | 2.449 | 2.226 | 2.671 | 33 |
| dprlo-hi80Hz | 1.861 | 1.585 | 2.136 | 33 |
| meanRTmslo20Hz | 366.053 | 348.407 | 383.699 | 33 |
| meanRTmshi20Hz | 481.749 | 462.601 | 500.898 | 33 |
| meanRTmslo-hi20Hz | -115.696 | -133.520 | -97.872 | 33 |
| meanRTmslo40Hz | 371.402 | 352.802 | 390.002 | 33 |
| meanRTmshi40Hz | 489.349 | 467.869 | 510.830 | 33 |
| meanRTmslo-hi40Hz | -117.947 | -135.053 | -100.841 | 33 |
| meanRTmslo80Hz | 366.977 | 353.100 | 380.853 | 33 |
| meanRTmshi80Hz | 489.550 | 470.105 | 508.996 | 33 |
| meanRTmslo-hi80Hz | -122.574 | -139.461 | -105.687 | 33 |
figure
Dtmp = subset(D, select=c('dprlo-hi20Hz', 'dprlo-hi40Hz', 'dprlo-hi80Hz'))
myylim = c(min(Dtmp), max(Dtmp))
RmCI = tableCIs(Dtmp)
#RmCI
Ddpr = RmCI[1:3,]
Dtmp = cbind(D$fp, Dtmp)
Dtmp = na.omit(Dtmp)
colnames(Dtmp) = c('fp','lo-hi20Hz','lo-hi40Hz','lo-hi80Hz')
bp1 = plotoverlay(dm = Dtmp, lbl = "d´", dvy = "Sensitivity (d´)", myylim)
Dtmp = subset(D, select=c('meanRTmslo-hi20Hz', 'meanRTmslo-hi40Hz', 'meanRTmslo-hi80Hz'))
myylim = c(min(Dtmp), max(Dtmp))
RmCI = tableCIs(Dtmp)
#RmCI
DRT = RmCI[1:3,1:4]
Dtmp = cbind(D$fp, Dtmp)
Dtmp = na.omit(Dtmp)
colnames(Dtmp) = c('fp','lo-hi20Hz','lo-hi40Hz','lo-hi80Hz')
bp2 = plotoverlay(dm = Dtmp, lbl = "RT", dvy = "RT (ms)", myylim)
grid.arrange(bp1, bp2, ncol = 2)
Ddpr$Condition = c("20 Hz", "40 Hz", "80 Hz")
DRT$Condition = c("20 Hz", "40 Hz", "80 Hz")
Ddpr$Condition = factor(Ddpr$Condition, levels=c("20 Hz", "40 Hz", "80 Hz"))
DRT$Condition = factor(DRT$Condition, levels=c("20 Hz", "40 Hz", "80 Hz"))
fig_dpr = ggplot(Ddpr, aes(x=Condition, y=Mean, fill=Condition)) +
geom_bar(stat="identity"
,color="black" # add black border to each bar
,position=position_dodge(), show.legend=FALSE) + # separare bars
theme_bw() + # get rid of background
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank()) + #no grid lines
geom_hline(yintercept = 0) +
geom_errorbar(position=position_dodge(.9), width=.25,
aes(ymin=LL, ymax=UL)) +
scale_fill_manual(
values = c("#999999", "#CCCCCC", "#FFFFFF")) +
labs(title = paste0("d´(N = ", nrow(D),")")) +
theme(plot.title = element_text(hjust = 0.5)) +
labs(x = "Low minus high load") +
labs(y = "Sensitivity (d´)")
fig_RT = ggplot(DRT, aes(x=Condition, y=Mean, fill=Condition)) +
geom_bar(stat="identity"
,color="black" # add black border to each bar
,position=position_dodge(), show.legend=FALSE) + # separare bars
theme_bw() + # get rid of background
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank()) + #no grid lines
geom_hline(yintercept = 0) +
geom_errorbar(position=position_dodge(.9), width=.25,
aes(ymin=LL, ymax=UL)) +
scale_fill_manual(
values = c("#999999", "#CCCCCC", "#FFFFFF")) +
labs(title = paste0("RT (N = ", nrow(D),")")) +
theme(plot.title = element_text(hjust = 0.5)) +
labs(x = "Low minus high load") +
labs(y = "Reaction time (ms)")
grid.arrange(fig_dpr, fig_RT, ncol = 2)
ANOVA
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: low and high
modulation frequency: 20, 40, 80 Hz
Dtmp1a = subset(D, select=c(fp, dprlo20Hz, dprlo40Hz, dprlo80Hz,
dprhi20Hz, dprhi40Hz, dprhi80Hz))
Dtmp1b = subset(D, select=c(fp, meanRTmslo20Hz, meanRTmslo40Hz, meanRTmslo80Hz,
meanRTmshi20Hz, meanRTmshi40Hz, meanRTmshi80Hz))
Dtmp1 = list(Dtmp1a, Dtmp1b)
Dvars = c("d´",'RT')
Fres = (matrix(ncol = 4, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','Overall_Interaction_p','BF01_Interaction', 'BF01_error%')d prime
## =====================================================================
## d´
## =====================================================================
##
##
## Descriptives
## ============
## load frequency N Mean CI_LL CI_UL
## low 20 33 4.370138 4.073691 4.666584
## low 40 33 4.354547 4.110703 4.598391
## low 80 33 4.309363 4.014338 4.604388
## high 20 33 2.482509 2.219678 2.745340
## high 40 33 2.569213 2.344854 2.793571
## high 80 33 2.448713 2.226218 2.671207
##
##
## Anova
## =====
##
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
##
## Sum Sq num Df Error SS den Df F value Pr(>F)
## (Intercept) 2319.16 1 61.248 32 1211.6901 <2e-16 ***
## load 168.41 1 15.088 32 357.1987 <2e-16 ***
## frequency 0.23 2 13.365 64 0.5459 0.5820
## load:frequency 0.09 2 12.946 64 0.2293 0.7958
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Mauchly Tests for Sphericity
##
## Test statistic p-value
## frequency 0.98458 0.78588
## load:frequency 0.78056 0.02149
##
##
## Greenhouse-Geisser and Huynh-Feldt Corrections
## for Departure from Sphericity
##
## GG eps Pr(>F[GG])
## frequency 0.98481 0.5794
## load:frequency 0.82005 0.7518
##
## HF eps Pr(>F[HF])
## frequency 1.0488947 0.5820013
## load:frequency 0.8584233 0.7621508
##
##
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + frequency + fp : 9.245612 ±3.19%
##
## Against denominator:
## dv ~ load + frequency + load:frequency + fp
## ---
## Bayes factor type: BFlinearModel, JZSRT (ms)
## =====================================================================
## RT
## =====================================================================
##
##
## Descriptives
## ============
## load frequency N Mean CI_LL CI_UL
## low 20 33 366.0533 348.4075 383.6991
## low 40 33 371.4023 352.8021 390.0024
## low 80 33 366.9766 353.1000 380.8532
## high 20 33 481.7493 462.6009 500.8978
## high 40 33 489.3494 467.8692 510.8297
## high 80 33 489.5503 470.1049 508.9957
##
##
## Anova
## =====
##
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
##
## Sum Sq num Df Error SS den Df F value Pr(>F)
## (Intercept) 36188029 1 388803 32 2978.4135 <2e-16 ***
## load 697898 1 93164 32 239.7144 <2e-16 ***
## frequency 1439 2 20541 64 2.2418 0.1145
## load:frequency 406 2 20794 64 0.6244 0.5388
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Mauchly Tests for Sphericity
##
## Test statistic p-value
## frequency 0.8463 0.07527
## load:frequency 0.9727 0.65112
##
##
## Greenhouse-Geisser and Huynh-Feldt Corrections
## for Departure from Sphericity
##
## GG eps Pr(>F[GG])
## frequency 0.86678 0.1225
## load:frequency 0.97342 0.5347
##
## HF eps Pr(>F[HF])
## frequency 0.912019 0.1197622
## load:frequency 1.035598 0.5387909
##
##
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + frequency + fp : 9.085288 ±2.34%
##
## Against denominator:
## dv ~ load + frequency + load:frequency + fp
## ---
## Bayes factor type: BFlinearModel, JZSANOVA summary
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: low and high
modulation frequency: 20, 40, 80 Hz
An interaction would suggest that load effects differed among the modulation frequencies.
Summary of interaction analyses:
* Overall_interaction has 2 dfs.
* 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.
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 | Overall_Interaction_p | BF01_Interaction | BF01_error% |
|---|---|---|---|
| d´ | 0.752 | 9.246 | 3.192 |
| RT | 0.535 | 9.085 | 2.337 |
visual P3
visual P3 = mean amplitude (between 300 and 400 ms) to targets minus nontargets
Note: Because there was no task during no load, there were no targets and nontargets to examine the P3.
table
Dtmp = subset(D, select=c('vP3lo20', 'vP3hi20', 'vP3lo-hi20',
'vP3lo40', 'vP3hi40', 'vP3lo-hi40',
'vP3lo80', 'vP3hi80', 'vP3lo-hi80'))
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 |
|---|---|---|---|---|
| vP3lo20 | 9.566 | 7.792 | 11.339 | 33 |
| vP3hi20 | 3.564 | 2.233 | 4.894 | 33 |
| vP3lo-hi20 | 6.002 | 4.247 | 7.757 | 33 |
| vP3lo40 | 9.700 | 8.227 | 11.172 | 33 |
| vP3hi40 | 3.290 | 2.001 | 4.579 | 33 |
| vP3lo-hi40 | 6.410 | 5.276 | 7.543 | 33 |
| vP3lo80 | 9.974 | 8.401 | 11.547 | 33 |
| vP3hi80 | 3.891 | 2.834 | 4.949 | 33 |
| vP3lo-hi80 | 6.083 | 4.634 | 7.531 | 33 |
figure
Dtmp = subset(D, select=c('vP3lo-hi20', 'vP3lo-hi40', 'vP3lo-hi80'))
RmCI = tableCIs(Dtmp)
myylim = c(min(Dtmp), max(Dtmp))
Dtmp = cbind(D$fp, Dtmp)
Dtmp = na.omit(Dtmp)
colnames(Dtmp) = c('fp','lo-hi20Hz','lo-hi40Hz','lo-hi80Hz')
bp1 = plotoverlay(dm = Dtmp,lbl = "Visual P3", dvy = "Mean amplitude (µV)", myylim)
plot(bp1)
RmCI$Variable = c("20 Hz", "40 Hz", "80 Hz")
fig_P3 = ggplot(RmCI, aes(x=Variable, y=Mean, fill=Variable)) +
geom_bar(stat="identity"
,color="black" # add black border to each bar
,position=position_dodge(), show.legend=TRUE) + # separare bars
theme_bw() + # get rid of background
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank()) + #no grid lines
geom_hline(yintercept = 0) +
geom_errorbar(position=position_dodge(.9), width=.25,
aes(ymin=LL, ymax=UL)) +
scale_fill_manual(
values = c("#999999", "#CCCCCC", "#FFFFFF"),
name = "Frequency") +
labs(title = paste0("Visual P3 (N = ", nrow(D),")")) +
theme(plot.title = element_text(hjust = 0.5)) +
labs(x = "Low minus high load") +
labs(y = "Mean amplitude (µV)")
plot(fig_P3)
ANOVA
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: low and high
modulation frequency: 20, 40, 80 Hz
Dtmp1a = subset(D, select=c(fp, vP3lo20, vP3lo40, vP3lo80,
vP3hi20, vP3hi40, vP3hi80))
# rename because Hz is missing at the end
colnames(Dtmp1a) = c('fp', 'vP3lo20Hz', 'vP3lo40Hz', 'vP3lo80Hz',
'vP3hi20Hz', 'vP3hi40Hz', 'vP3hi80Hz')
Dtmp1 = list(Dtmp1a)
Dvars = c("Visual P3")
Fres = (matrix(ncol = 4, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','Overall_Interaction_p','BF01_Interaction', 'BF01_error%')
Fres = computeANOVA(1, Dtmp1[[1]], mylvls = c('low','high'))
## =====================================================================
## Visual P3
## =====================================================================
##
##
## Descriptives
## ============
## load frequency N Mean CI_LL CI_UL
## low 20 33 9.565664 7.791973 11.339356
## low 40 33 9.699620 8.227182 11.172059
## low 80 33 9.974289 8.401440 11.547139
## high 20 33 3.563677 2.232946 4.894407
## high 40 33 3.290102 2.000892 4.579313
## high 80 33 3.891436 2.833961 4.948910
##
##
## Anova
## =====
##
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
##
## Sum Sq num Df Error SS den Df F value Pr(>F)
## (Intercept) 8793.3 1 2085.96 32 134.8951 5.158e-13 ***
## load 1881.2 1 520.63 32 115.6270 3.726e-12 ***
## frequency 7.3 2 232.13 64 1.0077 0.3708
## load:frequency 1.5 2 301.73 64 0.1629 0.8500
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Mauchly Tests for Sphericity
##
## Test statistic p-value
## frequency 0.75591 0.013072
## load:frequency 0.83321 0.059111
##
##
## Greenhouse-Geisser and Huynh-Feldt Corrections
## for Departure from Sphericity
##
## GG eps Pr(>F[GG])
## frequency 0.80380 0.3571
## load:frequency 0.85705 0.8176
##
## HF eps Pr(>F[HF])
## frequency 0.8398650 0.3599198
## load:frequency 0.9008364 0.8283076
##
##
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + frequency + fp : 9.775045 ±2.63%
##
## Against denominator:
## dv ~ load + frequency + load:frequency + fp
## ---
## Bayes factor type: BFlinearModel, JZSANOVA summary
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: low and high
modulation frequency: 20, 40, 80 Hz
An interaction would suggest that load effects differed among the modulation frequencies.
Summary of interaction analyses:
* Overall_interaction has 2 dfs.
* 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.
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 | Overall_Interaction_p | BF01_Interaction | BF01_error% |
|---|---|---|---|
| Visual P3 | 0.818 | 9.775 | 2.629 |
workload ratings (NASA-TLX)
workload ratings = mental demand and effort
table
Dnasa = data.frame()
tmpRmCI = data.frame(row.names = NULL)[1:18, ]
for (c in c('Mental', 'Effort')){
Dtmp = getdiffs(D, iv = c)
Dtmp = Dtmp[,-c(1:9)]
myylim = c(min(Dtmp), max(Dtmp))
RmCI = tableCIs(Dtmp)
#write.table(RmCI, file = paste0("Table_descr_",c,"_",nowsample,"_subjects.txt"),
# row.names = F, dec = ".", sep = "\t")
tmpRmCI = cbind(tmpRmCI, RmCI)
Dtmp2 = RmCI
Dtmp2$var = c
Dnasa = rbind(Dnasa, Dtmp2)
Dtmp = cbind(D$fp, Dtmp)
Dtmp = na.omit(Dtmp)
colnames(Dtmp)[1] = 'fp'
#bp1 = plotoverlay(dm = Dtmp, lbl = paste0("NASA ", c), dvy = "Mean rating", myylim)
#plot(bp1)
}
rownames(tmpRmCI) = NULL
tmpRmCI = tmpRmCI[,-c(5,6)]
tmpRmCI = subset(tmpRmCI, select=c(Variable, N, Mean, LL, UL, Mean.1, LL.1, UL.1))
tmpRmCI %>%
kable(digits = 3,
#caption = paste0("Workload ratings (N = ", nrow(D),")"),
col.names = c('Condition','N','Mean','LL','UL','Mean','LL','UL')) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left') %>%
add_header_above(c(" " = 2, "Mental demand" = 3, "Effort" = 3))| Condition | N | Mean | LL | UL | Mean | LL | UL |
|---|---|---|---|---|---|---|---|
| no20Hz | 33 | 16.561 | 11.272 | 21.849 | 15.167 | 9.666 | 20.667 |
| lo20Hz | 33 | 25.485 | 20.072 | 30.897 | 25.818 | 20.091 | 31.545 |
| hi20Hz | 33 | 38.864 | 32.794 | 44.933 | 40.348 | 33.873 | 46.824 |
| no-lo20Hz | 33 | -8.924 | -14.027 | -3.822 | -10.652 | -15.918 | -5.385 |
| no-hi20Hz | 33 | -22.303 | -28.381 | -16.225 | -25.182 | -32.000 | -18.364 |
| lo-hi20Hz | 33 | -13.379 | -17.596 | -9.161 | -14.530 | -20.882 | -8.179 |
| no40Hz | 33 | 15.318 | 10.419 | 20.218 | 13.273 | 8.768 | 17.778 |
| lo40Hz | 33 | 24.667 | 19.693 | 29.640 | 25.833 | 19.621 | 32.046 |
| hi40Hz | 33 | 38.606 | 33.052 | 44.160 | 42.167 | 36.613 | 47.721 |
| no-lo40Hz | 33 | -9.348 | -13.685 | -5.012 | -12.561 | -18.071 | -7.050 |
| no-hi40Hz | 33 | -23.288 | -28.837 | -17.738 | -28.894 | -33.968 | -23.819 |
| lo-hi40Hz | 33 | -13.939 | -17.357 | -10.522 | -16.333 | -21.252 | -11.415 |
| no80Hz | 33 | 14.470 | 9.980 | 18.960 | 13.212 | 9.094 | 17.330 |
| lo80Hz | 33 | 26.500 | 21.268 | 31.732 | 29.030 | 22.272 | 35.789 |
| hi80Hz | 33 | 39.348 | 32.801 | 45.896 | 41.939 | 35.256 | 48.623 |
| no-lo80Hz | 33 | -12.030 | -16.096 | -7.965 | -15.818 | -20.678 | -10.958 |
| no-hi80Hz | 33 | -24.879 | -31.866 | -17.892 | -28.727 | -35.584 | -21.870 |
| lo-hi80Hz | 33 | -12.848 | -17.441 | -8.256 | -12.909 | -18.676 | -7.142 |
colnames(Dnasa)[1] = 'Condition'
colnames(Dnasa)[6] = 'Variable'
tmp = Dnasa$Condition %in% c('no20Hz', 'lo20Hz', 'hi20Hz',
'no40Hz', 'lo40Hz', 'hi40Hz',
'no80Hz', 'lo80Hz', 'hi80Hz')
Dnasa = Dnasa[tmp,]
Dnasa$Condition = factor(Dnasa$Condition,
levels=c('no20Hz', 'lo20Hz', 'hi20Hz',
'no40Hz', 'lo40Hz', 'hi40Hz',
'no80Hz', 'lo80Hz', 'hi80Hz'))
Dnasa$Variable = factor(Dnasa$Variable, levels=c("Mental", "Effort"))
rm(c, bp1, bp2, tmp, Dtmp, Dtmp2, tmpRmCI, RmCI)figure
colnames(Dnasa)[1] = 'Cond'
Dnasa$Freq = NA
Dnasa$Freq[grepl(pattern = "20", Dnasa$Cond)] = 20
Dnasa$Freq[grepl(pattern = "40", Dnasa$Cond)] = 40
Dnasa$Freq[grepl(pattern = "80", Dnasa$Cond)] = 80
Dnasa$Freq = factor(Dnasa$Freq)
Dnasa$Condition = NA
Dnasa$Condition[grepl(pattern = "no", Dnasa$Cond)] = 'no'
Dnasa$Condition[grepl(pattern = "lo", Dnasa$Cond)] = 'low'
Dnasa$Condition[grepl(pattern = "hi", Dnasa$Cond)] = 'high'
Dnasa$Condition = factor(Dnasa$Condition,
levels=c('no', 'low', 'high'))
Dtmp = Dnasa[Dnasa$Variable=='Mental',]
fig_mental = ggplot(Dtmp, aes(x=Freq, y=Mean, fill=Condition)) +
geom_bar(stat="identity"
,color="black" # add black border to each bar
,position=position_dodge(), show.legend=TRUE) + # separate bars
theme_bw() + # get rid of background
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank()) + #no grid lines
geom_hline(yintercept = 0) +
geom_errorbar(position=position_dodge(.9), width=.25,
aes(ymin=LL, ymax=UL)) +
scale_fill_manual(
values = c("#FFFFFF", "#CCCCCC", "#999999")) +
labs(title = paste0("Mental demand (N = ", nrow(D),")")) +
theme(plot.title = element_text(hjust = 0.5)) +
labs(x = "Frequency (Hz)") +
labs(y = "Mean rating") +
ylim(0, 50)
Dtmp = Dnasa[Dnasa$Variable=='Effort',]
fig_effort = ggplot(Dtmp, aes(x=Freq, y=Mean, fill=Condition)) +
geom_bar(stat="identity"
,color="black" # add black border to each bar
,position=position_dodge(), show.legend=TRUE) + # separate bars
theme_bw() + # get rid of background
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank()) + #no grid lines
geom_hline(yintercept = 0) +
geom_errorbar(position=position_dodge(.9), width=.25,
aes(ymin=LL, ymax=UL)) +
scale_fill_manual(
values = c("#FFFFFF", "#CCCCCC", "#999999")) +
labs(title = paste0("Effort (N = ", nrow(D),")")) +
theme(plot.title = element_text(hjust = 0.5)) +
labs(x = "Frequency (Hz)") +
labs(y = "Mean rating") +
ylim(0, 50)
grid.arrange(fig_mental, fig_effort, ncol = 2)
bp1 = arrangeGrob(fig_mental, fig_effort, ncol = 2)
filename = "figure_NASA_ratings"
ggsave(file.path(dir_fig,paste0(filename,'.jpg')), bp1, dpi = 600, width = 7, height = 4)
ggsave(file.path(dir_fig,paste0(filename,'.pdf')), bp1, dpi = 600, width = 7, height = 4)
rm(bp1, Dnasa, Dtmp, fig_mental, fig_effort, filename)ANOVA
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: no, low, high
modulation frequency: 20, 40, 80 Hz
Dtmp1a = subset(D, select=c(fp, Mentalno20Hz, Mentalno40Hz, Mentalno80Hz,
Mentallo20Hz, Mentallo40Hz, Mentallo80Hz,
Mentalhi20Hz, Mentalhi40Hz, Mentalhi80Hz))
Dtmp1b = subset(D, select=c(fp, Effortno20Hz, Effortno40Hz, Effortno80Hz,
Effortlo20Hz, Effortlo40Hz, Effortlo80Hz,
Efforthi20Hz, Efforthi40Hz, Efforthi80Hz))
Dtmp1 = list(Dtmp1a, Dtmp1b)
Dvars = c("Mental", "Effort")
Fres = (matrix(ncol = 4, nrow = length(Dvars)))
Fres = as.data.frame(Fres)
colnames(Fres) = c('Variable','Overall_Interaction_p','BF01_Interaction', 'BF01_error%')Mental demand
## =====================================================================
## Mental
## =====================================================================
##
##
## Descriptives
## ============
## load frequency N Mean CI_LL CI_UL
## no 20 33 16.56061 11.272442 21.84877
## no 40 33 15.31818 10.418564 20.21780
## no 80 33 14.46970 9.979715 18.95968
## low 20 33 25.48485 20.072443 30.89725
## low 40 33 24.66667 19.693455 29.63988
## low 80 33 26.50000 21.268000 31.73200
## high 20 33 38.86364 32.794406 44.93287
## high 40 33 38.60606 33.052262 44.15986
## high 80 33 39.34848 32.800738 45.89623
##
##
## Anova
## =====
##
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
##
## Sum Sq num Df Error SS den Df F value Pr(>F)
## (Intercept) 210880 1 45936 32 146.9049 1.684e-13 ***
## load 27491 2 13238 64 66.4565 2.404e-16 ***
## frequency 32 2 1894 64 0.5394 0.5857
## load:frequency 106 4 6146 128 0.5525 0.6975
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Mauchly Tests for Sphericity
##
## Test statistic p-value
## load 0.49314 0.00002
## frequency 0.94958 0.44848
## load:frequency 0.49458 0.01105
##
##
## Greenhouse-Geisser and Huynh-Feldt Corrections
## for Departure from Sphericity
##
## GG eps Pr(>F[GG])
## load 0.66363 1.279e-11 ***
## frequency 0.95200 0.5772
## load:frequency 0.79558 0.6579
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## HF eps Pr(>F[HF])
## load 0.6813812 7.194686e-12
## frequency 1.0106339 5.857432e-01
## load:frequency 0.8936945 6.780188e-01
##
##
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + frequency + fp : 34.90733 ±7.07%
##
## Against denominator:
## dv ~ load + frequency + load:frequency + fp
## ---
## Bayes factor type: BFlinearModel, JZSEffort
## =====================================================================
## Effort
## =====================================================================
##
##
## Descriptives
## ============
## load frequency N Mean CI_LL CI_UL
## no 20 33 15.16667 9.666378 20.66696
## no 40 33 13.27273 8.767643 17.77781
## no 80 33 13.21212 9.094360 17.32988
## low 20 33 25.81818 20.090924 31.54544
## low 40 33 25.83333 19.621037 32.04563
## low 80 33 29.03030 22.272097 35.78851
## high 20 33 40.34848 33.873449 46.82352
## high 40 33 42.16667 36.612530 47.72080
## high 80 33 41.93939 35.256192 48.62260
##
##
## Anova
## =====
##
## Univariate Type III Repeated-Measures ANOVA Assuming Sphericity
##
## Sum Sq num Df Error SS den Df F value Pr(>F)
## (Intercept) 223316 1 47974 32 148.9565 1.402e-13 ***
## load 37751 2 18493 64 65.3255 3.480e-16 ***
## frequency 61 2 3545 64 0.5487 0.5804
## load:frequency 311 4 6854 128 1.4540 0.2201
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Mauchly Tests for Sphericity
##
## Test statistic p-value
## load 0.93184 0.33478
## frequency 0.97796 0.70789
## load:frequency 0.86374 0.87918
##
##
## Greenhouse-Geisser and Huynh-Feldt Corrections
## for Departure from Sphericity
##
## GG eps Pr(>F[GG])
## load 0.93619 2.67e-15 ***
## frequency 0.97843 0.5767
## load:frequency 0.93361 0.2233
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## HF eps Pr(>F[HF])
## load 0.9922499 4.456419e-16
## frequency 1.0414461 5.803804e-01
## load:frequency 1.0722973 2.200817e-01
##
##
## BF without vs with interaction
## ==============================
## Bayes factor analysis
## --------------
## [1] load + frequency + fp : 18.13436 ±2.66%
##
## Against denominator:
## dv ~ load + frequency + load:frequency + fp
## ---
## Bayes factor type: BFlinearModel, JZSANOVA summary
Analyze the factorial design (load x frequency) with repeated-measures ANOVA:
load: no, low, high
modulation frequency: 20, 40, 80 Hz
An interaction would suggest that load effects differed among the modulation frequencies.
Summary of interaction analyses:
* Overall_interaction has 4 dfs.
* 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.
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 | Overall_Interaction_p | BF01_Interaction | BF01_error% |
|---|---|---|---|
| Mental | 0.658 | 34.907 | 7.075 |
| Effort | 0.223 | 18.134 | 2.659 |
ASSR means and prereg BFs
Means and preregistered Bayes factors.
Ampl SmN
Amplitude signal minus noise (SmN)
plot
iv = "AmpSmN"
Dtmp = getdiffs(D, iv = iv)
myylim = c(min(Dtmp), max(Dtmp))
Dtmp = Dtmp[,c(10:27)]
Dtmp = cbind(D$fp, Dtmp)
colnames(Dtmp)[1] = 'fp'
Dtmp1 = Dtmp[,c(1,2:7)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1a = plotoverlay(dm = Dtmp1,lbl = paste0("20 Hz"), dvy = "Mean SmN amplitude (µV)", myylim)
Dtmp1 = Dtmp[,c(1,8:13)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1b = plotoverlay(dm = Dtmp1,lbl = paste0("40 Hz"), dvy = "", myylim)
Dtmp1 = Dtmp[,c(1,14:19)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1c = plotoverlay(dm = Dtmp1,lbl = paste0("80 Hz"), dvy = "", myylim)
grid.arrange(bp1a, bp1b, bp1c, ncol = 3)
table
iv = "AmpSmN"
Dtmp = getdiffs(D, iv = iv)
Dtmp = tableCIs(Dtmp[,-c(1:9)])
Dtmp %>%
kable(digits = 3) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left')| Variable | Mean | LL | UL | N |
|---|---|---|---|---|
| no20Hz | 0.360 | 0.272 | 0.447 | 33 |
| lo20Hz | 0.346 | 0.254 | 0.437 | 33 |
| hi20Hz | 0.355 | 0.254 | 0.455 | 33 |
| no-lo20Hz | 0.014 | -0.038 | 0.066 | 33 |
| no-hi20Hz | 0.005 | -0.062 | 0.072 | 33 |
| lo-hi20Hz | -0.009 | -0.060 | 0.043 | 33 |
| no40Hz | 0.479 | 0.388 | 0.571 | 33 |
| lo40Hz | 0.458 | 0.374 | 0.542 | 33 |
| hi40Hz | 0.442 | 0.351 | 0.532 | 33 |
| no-lo40Hz | 0.021 | -0.006 | 0.049 | 33 |
| no-hi40Hz | 0.038 | -0.003 | 0.078 | 33 |
| lo-hi40Hz | 0.017 | -0.017 | 0.050 | 33 |
| no80Hz | 0.099 | 0.064 | 0.133 | 33 |
| lo80Hz | 0.103 | 0.071 | 0.135 | 33 |
| hi80Hz | 0.092 | 0.057 | 0.128 | 33 |
| no-lo80Hz | -0.004 | -0.023 | 0.015 | 33 |
| no-hi80Hz | 0.006 | -0.016 | 0.028 | 33 |
| lo-hi80Hz | 0.011 | -0.012 | 0.033 | 33 |
preregistered BFs
The BFs are computed from Bayesian one-sample t tests of difference scores (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.
Dtmp = contrastBFs(D, iv = iv)
Dtmp %>%
kable(digits = 3,
col.names = c('Variable','Mean','[-1, +1]','[0, +1]','[0, +0.24]','[0, +0.37]')) %>%
add_header_above(c(" " = 2, "BF01" = 4)) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left')| Variable | Mean | [-1, +1] | [0, +1] | [0, +0.24] | [0, +0.37] |
|---|---|---|---|---|---|
| no-lo20Hz | 0.014 | 26.618 | 18.739 | 4.529 | 6.974 |
| no-hi20Hz | 0.005 | 23.783 | 21.015 | 5.084 | 7.827 |
| lo-hi20Hz | -0.009 | 29.386 | 39.497 | 9.624 | 14.798 |
| no-lo40Hz | 0.021 | 17.392 | 9.249 | 2.227 | 3.432 |
| no-hi40Hz | 0.038 | 6.995 | 3.616 | 0.869 | 1.339 |
| lo-hi40Hz | 0.016 | 28.865 | 17.013 | 4.109 | 6.328 |
| no-lo80Hz | -0.004 | 75.729 | 110.887 | 27.795 | 42.530 |
| no-hi80Hz | 0.006 | 61.778 | 42.256 | 10.305 | 15.843 |
| lo-hi80Hz | 0.011 | 44.927 | 26.853 | 6.510 | 10.018 |
ITC SmN
Intertrial coherence signal minus noise (SmN)
plot
iv = "ItcSmN"
Dtmp = getdiffs(D, iv = iv)
myylim = c(min(Dtmp), max(Dtmp))
Dtmp = Dtmp[,c(10:27)]
Dtmp = cbind(D$fp, Dtmp)
colnames(Dtmp)[1] = 'fp'
Dtmp1 = Dtmp[,c(1,2:7)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1a = plotoverlay(dm = Dtmp1,lbl = paste0("20 Hz"), dvy = "Mean SmN ITC", myylim)
Dtmp1 = Dtmp[,c(1,8:13)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1b = plotoverlay(dm = Dtmp1,lbl = paste0("40 Hz"), dvy = "", myylim)
Dtmp1 = Dtmp[,c(1,14:19)]
colnames(Dtmp1) = c('fp', 'no', 'low', 'high', 'no-low', 'no-high', 'low-high')
bp1c = plotoverlay(dm = Dtmp1,lbl = paste0("80 Hz"), dvy = "", myylim)
grid.arrange(bp1a, bp1b, bp1c, ncol = 3)
table
iv = "ItcSmN"
Dtmp = getdiffs(D, iv = iv)
Dtmp = tableCIs(Dtmp[,-c(1:9)])
Dtmp %>%
kable(digits = 3) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left')| Variable | Mean | LL | UL | N |
|---|---|---|---|---|
| no20Hz | 0.177 | 0.131 | 0.223 | 33 |
| lo20Hz | 0.190 | 0.146 | 0.234 | 33 |
| hi20Hz | 0.187 | 0.136 | 0.238 | 33 |
| no-lo20Hz | -0.013 | -0.033 | 0.008 | 33 |
| no-hi20Hz | -0.010 | -0.040 | 0.020 | 33 |
| lo-hi20Hz | 0.003 | -0.023 | 0.029 | 33 |
| no40Hz | 0.370 | 0.310 | 0.431 | 33 |
| lo40Hz | 0.345 | 0.288 | 0.403 | 33 |
| hi40Hz | 0.326 | 0.261 | 0.390 | 33 |
| no-lo40Hz | 0.025 | 0.000 | 0.050 | 33 |
| no-hi40Hz | 0.045 | 0.017 | 0.072 | 33 |
| lo-hi40Hz | 0.020 | -0.011 | 0.050 | 33 |
| no80Hz | 0.127 | 0.085 | 0.169 | 33 |
| lo80Hz | 0.116 | 0.076 | 0.157 | 33 |
| hi80Hz | 0.103 | 0.067 | 0.138 | 33 |
| no-lo80Hz | 0.011 | -0.010 | 0.032 | 33 |
| no-hi80Hz | 0.025 | -0.001 | 0.051 | 33 |
| lo-hi80Hz | 0.014 | -0.010 | 0.038 | 33 |
preregistered BFs
The BFs are computed from Bayesian one-sample t tests of difference scores (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.
Dtmp = contrastBFs(D, iv = iv)
Dtmp %>%
kable(digits = 3,
col.names = c('Variable','Mean','[-1, +1]','[0, +1]','[0, +0.24]','[0, +0.37]')) %>%
add_header_above(c(" " = 2, "BF01" = 4)) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left')| Variable | Mean | [-1, +1] | [0, +1] | [0, +0.24] | [0, +0.37] |
|---|---|---|---|---|---|
| no-lo20Hz | -0.013 | 34.192 | 152.465 | 38.902 | 59.341 |
| no-hi20Hz | -0.010 | 42.804 | 82.271 | 20.388 | 31.258 |
| lo-hi20Hz | 0.003 | 61.133 | 50.238 | 12.291 | 18.885 |
| no-lo40Hz | 0.025 | 8.614 | 4.408 | 1.060 | 1.633 |
| no-hi40Hz | 0.045 | 0.424 | 0.212 | 0.051 | 0.079 |
| lo-hi40Hz | 0.020 | 22.157 | 12.197 | 2.941 | 4.530 |
| no-lo80Hz | 0.011 | 43.737 | 25.432 | 6.162 | 9.484 |
| no-hi80Hz | 0.025 | 10.113 | 5.207 | 1.252 | 1.929 |
| lo-hi80Hz | 0.014 | 33.898 | 19.191 | 4.639 | 7.143 |
Floor effects?
Does the combined effect of task and load (i.e., no load minus high load) for SmN depend on the initial SmN value for no load?
A floor effect would be supported if large difference scores of no minus high load (on Y axis) can be observed only for large initial values of no load (on X axis).
In the plots, the regression lines are derived from local polynomial regression fitting, see geom_smooth(method=‘loess’).
Amp SmN
Results suggest no obvious relationship.
Note: 40-Hz shows an increase, but this is driven by two data points.
# amp SmN
Dtmp = Dtmp_amp
tymin = min(Dtmp$'no-hi20Hz',Dtmp$'no-hi40Hz',Dtmp$'no-hi80Hz')
tymax = max(Dtmp$'no-hi20Hz',Dtmp$'no-hi40Hz',Dtmp$'no-hi80Hz')
bp1 = ggplot(Dtmp, aes(x=`no20Hz`, y=`no-hi20Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("20 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
bp2 = ggplot(Dtmp, aes(x=`no40Hz`, y=`no-hi40Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("40 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
bp3 = ggplot(Dtmp, aes(x=`no80Hz`, y=`no-hi80Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("80 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
grid.arrange(bp1, bp2, bp3, ncol = 3)
ITC SmN
Results suggest no obvious relationship.
Note: 80-Hz shows an increase, but this is driven by few data points.
# itc SmN
Dtmp = Dtmp_itc
tymin = min(Dtmp$'no-hi20Hz',Dtmp$'no-hi40Hz',Dtmp$'no-hi80Hz')
tymax = max(Dtmp$'no-hi20Hz',Dtmp$'no-hi40Hz',Dtmp$'no-hi80Hz')
bp1 = ggplot(Dtmp, aes(x=`no20Hz`, y=`no-hi20Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("20 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
bp2 = ggplot(Dtmp, aes(x=`no40Hz`, y=`no-hi40Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("40 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
bp3 = ggplot(Dtmp, aes(x=`no80Hz`, y=`no-hi80Hz`)) +
theme_bw() + # get rid of background
geom_point(shape=1, size = 4) + # Use hollow circles
labs(title = paste0("80 Hz (N = ",nrow(D),")")) +
labs(x = "No load SmN") +
labs(y = "No minus high SmN") +
geom_hline(yintercept = 0) +
ylim(tymin,tymax) +
geom_smooth(method = "loess", formula = y ~ x)
grid.arrange(bp1, bp2, bp3, ncol = 3)
Exploratory BFs
In the preregistered BF analyses, we defined upper limits of the priors on the basis of the mean SmN values for no load in a previous study with 40-Hz ASSRs (Study 2 in Szychowska & Wiens, 2020). Theoretically, the SmN during no load defines an upper limit of how strong the task effect and load effect can be.
Because the visual task in the present study was identical to that in the previous study, we anticipated that we would observe similar mean SmN values during no load for 40-Hz. So, we used our previous data to predict the upper limits for 40 Hz. For simplicity, we used these upper limits also for 20 and 80 Hz. Thus, the original goal was to define realistic upper limits in the BF analyses and to preregister these limits to facilitate confirmatory research (hypothesis testing).
However, the present results showed that the actual SmN during no and low load differed from the preregistered upper limits. To conduct BF analyses that matched our original goal of using reasonable upper limits, we defined the upper limits on the basis of the present data, separately for each measure and frequency.
No minus low/high
One analysis considered no load as the baseline, and the H1 model was defined as a uniform distribution between 0 (as lower limit, LL) and the present mean of no load (as upper limit, UL).
No minus low = task effect
No minus high = combined effect of task and load
Low minus high
The other analysis considered low load as the baseline, and the H1 model was defined as a uniform distribution between 0 (as lower limit, LL) and the present mean of low load (as upper limit, UL).
Low minus high = load effect
In the table, LL and UL refer to the 95% CIs of the means (and not to the LL and UL of the H1 models).
no minus low/high
# create matrix with measure, freq, mean no load,
# diff noMlo, LL_CI, UL_CI, BF01 noMlo,
# diff noMhi, LL_CI, UL_CI, BF01 noMhi
DBF01 = numeric()
Dtmp_freq = c(20,40,80)
# amp ----
Dtmp_H1 = subset(Dtmp_amp,
select=c(no20Hz, no40Hz, no80Hz))
Dtmp_noMlo = subset(Dtmp_amp,
select=c('no-lo20Hz', 'no-lo40Hz', 'no-lo80Hz'))
Dtmp_noMhi = subset(Dtmp_amp,
select=c('no-hi20Hz', 'no-hi40Hz', 'no-hi80Hz'))
LL = 0
for (i in 1:3) {
# define UL as mean of no load
UL = mean(Dtmp_H1[,i])
# noMlo
tmp = Dtmp_noMlo[,i]
# noMhi
tmp2 = Dtmp_noMhi[,i]
DBF01 = rbind(DBF01, c(Dtmp_freq[i], UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
}
rm(Dtmp_H1, Dtmp_noMlo, Dtmp_noMhi)
# itc ----
Dtmp_H1 = subset(Dtmp_itc,
select=c(no20Hz, no40Hz, no80Hz))
Dtmp_noMlo = subset(Dtmp_itc,
select=c('no-lo20Hz', 'no-lo40Hz', 'no-lo80Hz'))
Dtmp_noMhi = subset(Dtmp_itc,
select=c('no-hi20Hz', 'no-hi40Hz', 'no-hi80Hz'))
LL = 0
for (i in 1:3) {
# define UL as mean of no load
UL = mean(Dtmp_H1[,i])
# noMlo
tmp = Dtmp_noMlo[,i]
# noMhi
tmp2 = Dtmp_noMhi[,i]
DBF01 = rbind(DBF01, c(Dtmp_freq[i], UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
}
DBF01 = as.data.frame(DBF01)
colnames(DBF01)=c('Freq','M-no',
'M-noMlo','LL-noMlo','UL-noMlo','BF01-noMlo',
'M-noMhi','LL-noMhi','UL-noMhi','BF01-noMhi')
DBF01$Measure = c('Amp','Amp','Amp','ITC','ITC','ITC')
DBF01 = DBF01[,c(ncol(DBF01),1:ncol(DBF01)-1)]
DBF01 %>%
kable(digits = 3,
col.names = c('Measure','Frequency','Mean',
'Mean','LL', 'UL', 'BF01',
'Mean','LL', 'UL', 'BF01')) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left') %>%
add_header_above(c(" " = 2, "no" = 1, "no minus low" = 4, "no minus high" = 4))| Measure | Frequency | Mean | Mean | LL | UL | BF01 | Mean | LL | UL | BF01 |
|---|---|---|---|---|---|---|---|---|---|---|
| Amp | 20 | 0.360 | 0.014 | -0.038 | 0.066 | 6.778 | 0.005 | -0.062 | 0.072 | 7.607 |
| Amp | 40 | 0.479 | 0.021 | -0.006 | 0.049 | 4.444 | 0.038 | -0.003 | 0.078 | 1.735 |
| Amp | 80 | 0.099 | -0.004 | -0.023 | 0.015 | 11.529 | 0.006 | -0.016 | 0.028 | 4.252 |
| ITC | 20 | 0.177 | -0.013 | -0.033 | 0.008 | 28.834 | -0.010 | -0.040 | 0.020 | 15.074 |
| ITC | 40 | 0.370 | 0.025 | 0.000 | 0.050 | 1.635 | 0.045 | 0.017 | 0.072 | 0.079 |
| ITC | 80 | 0.127 | 0.011 | -0.010 | 0.032 | 3.274 | 0.025 | -0.001 | 0.051 | 0.665 |
low minus high
# create matrix with measure, freq, mean low load,
# diff loMhi, LL_CI, UL_CI, BF01 looMhi
DBF01 = numeric()
Dtmp_freq = c(20,40,80)
# amp ----
Dtmp_H1 = subset(Dtmp_amp,
select=c(lo20Hz, lo40Hz, lo80Hz))
Dtmp_loMhi = subset(Dtmp_amp,
select=c('lo-hi20Hz', 'lo-hi40Hz', 'lo-hi80Hz'))
LL = 0
for (i in 1:3) {
# define UL as mean of no load
UL = mean(Dtmp_H1[,i])
# loMhi
tmp = Dtmp_loMhi[,i]
DBF01 = rbind(DBF01, c(Dtmp_freq[i], UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
}
rm(Dtmp_H1, Dtmp_loMhi)
# itc ----
Dtmp_H1 = subset(Dtmp_itc,
select=c(lo20Hz, lo40Hz, lo80Hz))
Dtmp_loMhi = subset(Dtmp_itc,
select=c('lo-hi20Hz', 'lo-hi40Hz', 'lo-hi80Hz'))
LL = 0
for (i in 1:3) {
# define UL as mean of no load
UL = mean(Dtmp_H1[,i])
# loMhi
tmp = Dtmp_loMhi[,i]
DBF01 = rbind(DBF01, c(Dtmp_freq[i], UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
}
DBF01 = as.data.frame(DBF01)
colnames(DBF01)=c('Freq','M-no',
'M-loMhi','LL-loMhi','UL-loMhi','BF01-loMhi')
DBF01$Measure = c('Amp','Amp','Amp','ITC','ITC','ITC')
DBF01 = DBF01[,c(ncol(DBF01),1:ncol(DBF01)-1)]
DBF01 %>%
kable(digits = 3,
col.names = c('Measure','Frequency','Mean',
'Mean','LL', 'UL', 'BF01')) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left') %>%
add_header_above(c(" " = 2, "low" = 1, "low minus high" = 4))| Measure | Frequency | Mean | Mean | LL | UL | BF01 |
|---|---|---|---|---|---|---|
| Amp | 20 | 0.346 | -0.009 | -0.060 | 0.043 | 13.833 |
| Amp | 40 | 0.458 | 0.017 | -0.017 | 0.050 | 7.831 |
| Amp | 80 | 0.103 | 0.011 | -0.012 | 0.033 | 2.801 |
| ITC | 20 | 0.190 | 0.003 | -0.023 | 0.029 | 9.739 |
| ITC | 40 | 0.345 | 0.020 | -0.011 | 0.050 | 4.228 |
| ITC | 80 | 0.116 | 0.014 | -0.010 | 0.038 | 2.254 |
40-Hz: combine with previous data
Our previous study (Szychowska & Wiens, 2020) studied effects of task and load on 40-Hz ASSRs.
Exp. 1: Experiment 1 (N = 43) in Szychowska & Wiens (2020). This study included low and high load.
Exp. 2: Experiment 2 (N = 45) in Szychowska & Wiens (2020). This study included no, low, high, and very high load. Because results showed that performance was impaired more strongly during high load than very high load, high load was used below (and very high load was excluded).
Below, we combined the previous data (Exp. 1 and 2) with the present data to obtain an estimate on the basis of all of our data on 40-Hz ASSRs.
Dtmpa = subset(Dtmp_amp,
select=c(fp, no40Hz, lo40Hz, hi40Hz))
colnames(Dtmpa) = c('fp', 'AmpSmNno', 'AmpSmNlo', 'AmpSmNhi')
Dtmpi = subset(Dtmp_itc,
select=c(no40Hz, lo40Hz, hi40Hz))
colnames(Dtmpi) = c('ItcSmNno', 'ItcSmNlo', 'ItcSmNhi')
Dtmp = cbind(Dtmpa, Dtmpi)
Dtmp$study = 'present'
DS1=read.csv(file.path(dir_res,"data_export_SmN_ASSR_study1_all_subjects.tsv"), sep="\t",header=TRUE)
DS2=read.csv(file.path(dir_res,"data_export_SmN_ASSR_study2.tsv"), sep="\t",header=TRUE)no minus low/high
No load is used as the upper limit of the H1 model.
H1 model:
LL = 0
UL = no load
Note that Exp. 1 is excluded because no load was not administered.
No minus low = task effect
No minus high = combined effect of task and load
In the table, LL and UL refer to the 95% CIs of the means (and not to the LL and UL of the H1 models).
# compare noMlo and noMhi
Dall = merge(Dtmp, DS2, all = T)
DBF01 = numeric()
# create matrix with measure, study, n, mean no load,
# diff noMlo, LL 95%CI, UL 95%CI, BF01 noMlo,
# diff noMhi, LL 95%CI, UL 95%CI, BF01 noMhi
LL = 0
# amp
# study 2
UL = mean(DS2$AmpSmNno)
tmp = DS2$AmpSmNno-DS2$AmpSmNlo
tmp2 = DS2$AmpSmNno-DS2$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Exp. 2', nrow(DS2), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
# present study
UL = mean(Dtmp$AmpSmNno)
tmp = Dtmp$AmpSmNno-Dtmp$AmpSmNlo
tmp2 = Dtmp$AmpSmNno-Dtmp$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Present', nrow(Dtmp), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
# combined
UL = mean(Dall$AmpSmNno)
tmp = Dall$AmpSmNno-Dall$AmpSmNlo
tmp2 = Dall$AmpSmNno-Dall$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Combined', nrow(Dall), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
# itc
# study 2
UL = mean(DS2$ItcSmNno)
tmp = DS2$ItcSmNno-DS2$ItcSmNlo
tmp2 = DS2$ItcSmNno-DS2$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Exp. 2', nrow(DS2), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
# present study
UL = mean(Dtmp$ItcSmNno)
tmp = Dtmp$ItcSmNno-Dtmp$ItcSmNlo
tmp2 = Dtmp$ItcSmNno-Dtmp$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Present', nrow(Dtmp), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
# combined
UL = mean(Dall$ItcSmNno)
tmp = Dall$ItcSmNno-Dall$ItcSmNlo
tmp2 = Dall$ItcSmNno-Dall$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Combined', nrow(Dall), UL,
mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp),
mean(tmp2), ci(tmp2), BFfitH1(LL = 0, UL, tmp2)))
DBF01 = as.data.frame(DBF01)
colnames(DBF01)=c('Measure','Study','N','Mno',
'M-noMlo','LL-noMlo','UL-noMlo','BF01-noMlo',
'M-noMhi','LL-noMhi','UL-noMhi','BF01-noMhi')
# convert character to numeric
i = 3:12
DBF01[, i] = apply(DBF01[,i],2, function(x) as.numeric(as.character(x)))
DBF01 %>%
kable(digits = 3,
col.names = c('Measure','Study','N','Mean',
'Mean','LL', 'UL', 'BF01',
'Mean','LL', 'UL', 'BF01')) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left') %>%
add_header_above(c(" " = 3, "no" = 1, "no minus low" = 4, "no minus high" = 4)) %>%
row_spec(c(3,6), bold = T)| Measure | Study | N | Mean | Mean | LL | UL | BF01 | Mean | LL | UL | BF01 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Amp | Exp. 2 | 45 | 0.243 | -0.001 | -0.012 | 0.010 | 39.213 | 0.000 | -0.013 | 0.012 | 32.045 |
| Amp | Present | 33 | 0.479 | 0.021 | -0.006 | 0.049 | 4.444 | 0.038 | -0.003 | 0.078 | 1.735 |
| Amp | Combined | 78 | 0.343 | 0.008 | -0.005 | 0.022 | 10.065 | 0.016 | -0.003 | 0.034 | 3.785 |
| ITC | Exp. 2 | 45 | 0.368 | -0.004 | -0.021 | 0.012 | 51.978 | 0.005 | -0.015 | 0.024 | 19.749 |
| ITC | Present | 33 | 0.370 | 0.025 | 0.000 | 0.050 | 1.635 | 0.045 | 0.017 | 0.072 | 0.079 |
| ITC | Combined | 78 | 0.369 | 0.008 | -0.006 | 0.022 | 12.382 | 0.022 | 0.005 | 0.038 | 0.639 |
low minus high
Low load is used as the upper limit of the H1 model.
H1 model:
LL = 0
UL = low load
Low minus high = load effect
In the table, LL and UL refer to the 95% CIs of the means (and not to the LL and UL of the H1 models).
# compare low with high load
Dall = merge(Dtmp, DS1, all = T)
Dall = merge(Dall, DS2, all = T)
DBF01 = numeric()
# create matrix with measure, study, n, mean low load,
# diff loMhi, LL 95%CI, UL 95%CI, BF01 loMhi
LL = 0
# amp
# study 1
UL = mean(DS1$AmpSmNlo)
tmp = DS1$AmpSmNlo-DS1$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Exp. 1', nrow(DS1), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# study 2
UL = mean(DS2$AmpSmNlo)
tmp = DS2$AmpSmNlo-DS2$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Exp. 2', nrow(DS2), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# present study
UL = mean(Dtmp$AmpSmNlo)
tmp = Dtmp$AmpSmNlo-Dtmp$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Present', nrow(Dtmp), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# combined
UL = mean(Dall$AmpSmNlo)
tmp = Dall$AmpSmNlo-Dall$AmpSmNhi
DBF01 = rbind(DBF01, c('Amp','Combined', nrow(Dall), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# itc
# study 1
UL = mean(DS1$ItcSmNlo)
tmp = DS1$ItcSmNlo-DS1$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Exp. 1', nrow(DS1), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# study 2
UL = mean(DS2$ItcSmNlo)
tmp = DS2$ItcSmNlo-DS2$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Exp. 2', nrow(DS2), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# present study
UL = mean(Dtmp$ItcSmNlo)
tmp = Dtmp$ItcSmNlo-Dtmp$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Present', nrow(Dtmp), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
# combined
UL = mean(Dall$ItcSmNlo)
tmp = Dall$ItcSmNlo-Dall$ItcSmNhi
DBF01 = rbind(DBF01, c('ITC','Combined', nrow(Dall), UL, mean(tmp), ci(tmp), BFfitH1(LL = 0, UL, tmp)))
DBF01 = as.data.frame(DBF01)
colnames(DBF01)=c('Measure','Study','N','Mean','Mean','LL-95%CI', 'UL-95%CI', 'BF01')
# convert character to numeric
i = 3:8
DBF01[, i] = apply(DBF01[,i],2, function(x) as.numeric(as.character(x)))
DBF01 %>%
kable(digits = 3,
col.names = c('Measure','Study','N','Mean','Mean','LL', 'UL', 'BF01')) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = F, position = 'left') %>%
add_header_above(c(" " = 3, "low" = 1, "low minus high" = 4)) %>%
row_spec(c(4,8), bold = T)| Measure | Study | N | Mean | Mean | LL | UL | BF01 |
|---|---|---|---|---|---|---|---|
| Amp | Exp. 1 | 43 | 0.191 | 0.006 | -0.004 | 0.017 | 7.760 |
| Amp | Exp. 2 | 45 | 0.244 | 0.000 | -0.012 | 0.013 | 28.960 |
| Amp | Present | 33 | 0.458 | 0.017 | -0.017 | 0.050 | 7.831 |
| Amp | Combined | 121 | 0.283 | 0.007 | -0.004 | 0.018 | 10.069 |
| ITC | Exp. 1 | 43 | 0.309 | 0.019 | 0.000 | 0.038 | 1.924 |
| ITC | Exp. 2 | 45 | 0.373 | 0.009 | -0.009 | 0.027 | 11.840 |
| ITC | Present | 33 | 0.345 | 0.020 | -0.011 | 0.050 | 4.228 |
| ITC | Combined | 121 | 0.343 | 0.015 | 0.003 | 0.028 | 1.091 |