Skip to content

Commit

Permalink
(Re)run all logistic model versions. Add inline summary of stats
Browse files Browse the repository at this point in the history
  • Loading branch information
@l-acs
l-acs committed Dec 23, 2024
1 parent f81ceed commit 5424a12
Showing 1 changed file with 113 additions and 119 deletions.
232 changes: 113 additions & 119 deletions code/analysisWordLevelReadAloudBeta.R
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# readAloud-valence-beta Reading Task Analyses
# Authors: Luc Sahar, Jessica M. Alexander
# Last Updated: 2024-12-12
# Last Updated: 2024-12-23

# INPUTS
# data/df: behavioral data, for each participant on each passage, with relevant participant information and trial-level stimulus information
Expand Down Expand Up @@ -577,6 +577,7 @@ if (DEBUG) {
data=errorDat, REML=TRUE)
}

# n.s.
model2.5_z_scored <- glmer(misprod_outcome ~ scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(model2.5_z_scored)
Expand All @@ -588,100 +589,69 @@ if(DEBUG) { # compare to how it was/would've been before splitting outcome and p
summary(wrong_model2.5)
}

# tldr 2/28/24 SA indv. do not misproduce more/less


#hesitation x scaaredSoc
# model5 <- glmer(hesitation_outcome ~ scaaredSoc_gmc + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(model5)

# hesitation x scaaredSoc, control for word
model5.5_z_scored <- glmer(hesitation_outcome ~ scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(model5.5_z_scored)

# results are similar
# tldr 2/28/24: SA indv.s DO hesitate more

# revision 12/17/24: use the correct kind of model

# outcome is binary
# hesitation x scaaredSoc_z, control for word, binary outcome
# scaaredSoc_z 0.21491 0.07715 2.786 0.00534 **
model5.5_z_scored_logistic <- glmer(hesitation_outcome ~ scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(model5.5_z_scored_logistic)

# `REML` does not work with glmer, nor does `family` with lmer
# model5.5_z_scored_logistic_reml <- glmer(hesitation_outcome ~ scaaredSoc_z + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial", REML=TRUE)


#### supplemental analyses
# see notes

# Now, misproduction-hesitation relationships

# Errors as explained by disfluency: rate of misproduced syllables from rate of hesitated syllables
f_model20 <- glmer(misprod_outcome ~ hesitation_predictor + (1|id) + (1|passage),
# hesitation_predictor1 0.39511 0.03292 12.00 <2e-16 ***
f_model20_logistic <- glmer(misprod_outcome ~ hesitation_predictor + (1|id) + (1|passage),
data=errorDat, family = "binomial")
summary(f_model20) # ***
summary(f_model20_logistic)

# should we (1|word) here?
# no... that's not rate, is it? 3/27/24
# that's whether a word being hesitated on predicts it (that same word) being misproduced
# right?

# Errors as explained by disfluency: rate of misproduced words from rate of hesitated words
# f_model21 <- glmer(words_with_misprod ~ words_with_hes + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(f_model21) # ***
#
#
# # Errors as explained by disfluency: rate of misproduced words from rate of hesitated syllables
# f_model22 <- glmer(words_with_misprod ~ hesitation + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(f_model22) # ***
# commented out 2/28/24 bc word level isn't applicable here; _all_ of this is at
# word level
# NB my * comments here (this section of models at least) are out of date
# Errors as explained by disfluency: rate of misproduced syllables from rate of hesitated syllables, control for word
# hesitation_predictor1 0.09129 0.03670 2.488 0.0129 *
f_model20.5_logistic <- glmer(misprod_outcome ~ hesitation_predictor + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(f_model20.5_logistic)


# Now, misproduction-hesitation interactions with social anxiety

# Errors as explained by disfluency and SA: rate of misproduced syllables from rate of hesitated syllables and scaared
f_model23_z_scored <- glmer(misprod_outcome ~ hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage),
# hesitation_predictor1 0.396320 0.033051 11.991 <2e-16 ***
# scaaredSoc_z 0.006507 0.078505 0.083 0.934
# hesitation_predictor1:scaaredSoc_z -0.013847 0.031155 -0.444 0.657
f_model23_z_scored_logistic <- glmer(misprod_outcome ~ hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage),
data=errorDat, family = "binomial")
summary(f_model23_z_scored)

# Errors as explained by disfluency and SA: rate of misproduced words from rate of hesitated words and scaared
# f_model24 <- glmer(words_with_misprod ~ words_with_hes * scaaredSoc_gmc + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(f_model24)
# commented out 2/28/24 bc word level isn't applicable here; _all_ of this is at
# word level

summary(f_model23_z_scored_logistic)

# Errors as explained by disfluency and SA: misproduction from hesitation and scaared_z, control for word
# hesitation_predictor1 0.09261 0.03686 2.512 0.012 *
# scaaredSoc_z 0.01159 0.08326 0.139 0.889
# hesitation_predictor1:scaaredSoc_z -0.01278 0.03332 -0.384 0.701
f_model23.5_z_scored_logistic <- glmer(misprod_outcome ~ hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(f_model23.5_z_scored_logistic)

# Errors as explained by disfluency and SA: rate of misproduced words from rate of hesitated syllables and scaared
# f_model25 <- glmer(words_with_misprod ~ hesitation * scaaredSoc_gmc + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(f_model25)
# commented out 2/28/24 bc word level isn't applicable here; _all_ of this is at
# word level



# What happens when we control for age?
#hesitation x scaaredSoc
age_model1_z_scored <- glmer(hesitation_outcome ~ scaaredSoc_z + age_z + (1|id) + (1|passage),
# hesitation x scaaredSoc_z, cf. model5.5
# scaaredSoc_z 0.19651 0.07384 2.661 0.00779 **
# age_z 0.16195 0.07340 2.206 0.02737 *
age_model1_z_scored_logistic <- glmer(hesitation_outcome ~ scaaredSoc_z + age_z + (1|id) + (1|passage),
data=errorDat, family = "binomial")
summary(age_model1_z_scored)
summary(age_model1_z_scored_logistic)

# What happens when we control for age?
# hesitation x scaaredSoc_z, control for word, cf. model5.5
# scaaredSoc_z 0.20149 0.07361 2.737 0.0062 **
# age_z 0.16721 0.07311 2.287 0.0222 *
age_model1.5_z_scored_logistic <- glmer(hesitation_outcome ~ scaaredSoc_z + age_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(age_model1.5_z_scored_logistic)

#words_with_hes x scaaredSoc
# age_model2 <- glmer(words_with_hes ~ scaaredSoc_gmc + age_gmc + (1|id) + (1|passage),
# data=errorDat, family = "binomial")
# summary(age_model2)
# commented out 2/28/24 bc word level isn't applicable here; _all_ of this is at
# word level


# misprod-hes ordering
Expand Down Expand Up @@ -797,51 +767,51 @@ errorDat$log10frequency_with_absents_as_median_z <- scale(errorDat$log10frequenc
errorDat <- errorDat %>% select(-log10frequency_with_absents_as_median)

# Does a word's frequency predict hesitation on that word?
wordfreq_model_with_absents_as_median_1_z_scored <- glmer(hesitation_outcome ~ log10frequency_with_absents_as_median_z + (1|id) + (1|passage) + (1|word),
# log10frequency_with_absents_as_median_z -0.44253 0.04150 -10.66 <2e-16 ***
wordfreq_model_with_absents_as_median_1_z_scored_logistic <- glmer(hesitation_outcome ~ log10frequency_with_absents_as_median_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_1_z_scored)
summary(wordfreq_model_with_absents_as_median_1_z_scored_logistic)

# "" misprod on that word?
wordfreq_model_with_absents_as_median_2_z_scored <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z + (1|id) + (1|passage) + (1|word),
# Does a word's frequency predict misprod on that word?
# log10frequency_with_absents_as_median_z -0.68143 0.05109 -13.34 <2e-16 ***
wordfreq_model_with_absents_as_median_2_z_scored_logistic <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_2_z_scored)
summary(wordfreq_model_with_absents_as_median_2_z_scored_logistic)


# Do social anxiety and frequency interact to the presence of a hesitation or misproduction?
wordfreq_model_with_absents_as_median_3_z_scored <- glmer(hesitation_outcome ~ log10frequency_with_absents_as_median_z * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_3_z_scored)

# outcome is binary
# Do social anxiety and frequency interact to the presence of a hesitation?
# log10frequency_with_absents_as_median_z -0.444116 0.041562 -10.686 < 2e-16 ***
# scaaredSoc_z 0.217698 0.077599 2.805 0.00502 **
# log10frequency_with_absents_as_median_z:scaaredSoc_z 0.009726 0.013909 0.699 0.48441
wordfreq_model_with_absents_as_median_3_z_scored_logistic <- glmer(hesitation_outcome ~ log10frequency_with_absents_as_median_z * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family="binomial")
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_3_z_scored_logistic)

# wordfreq_model_with_absents_as_median_4_z_scored <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(wordfreq_model_with_absents_as_median_4_z_scored)

# Do social anxiety and frequency interact to the presence of a misproduction?
# log10frequency_with_absents_as_median_z -0.681946 0.051084 -13.349 <2e-16 ***
# scaaredSoc_z 0.007296 0.078917 0.092 0.9263
# log10frequency_with_absents_as_median_z:scaaredSoc_z -0.032234 0.016058 -2.007 0.0447 *
wordfreq_model_with_absents_as_median_4_z_scored_logistic <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_4_z_scored_logistic)

# Do frequency and the presence of a hesitation interact to predict the presence of a misproduction?
wordfreq_model_with_absents_as_median_5_z_scored <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * hesitation_predictor + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_5_z_scored)

# outcome is binary
# Do frequency and the presence of a hesitation interact to predict the presence of a misproduction?
# log10frequency_with_absents_as_median_z -0.73044 0.06304 -11.587 <2e-16 ***
# hesitation_predictor1 0.02546 0.05489 0.464 0.643
# log10frequency_with_absents_as_median_z:hesitation_predictor1 -0.05939 0.04113 -1.444 0.149
wordfreq_model_with_absents_as_median_5_z_scored_logistic <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * hesitation_predictor + (1|id) + (1|passage) + (1|word),
data=errorDat, family="binomial")
summary(wordfreq_model_with_absents_as_median_5_z_scored_logistic)

# Do frequency, SA, an d the presence of a hesitation all interact to predict the presence of a misproduction?
wordfreq_model_with_absents_as_median_6_z_scored <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(wordfreq_model_with_absents_as_median_6_z_scored)
# Do frequency, SA, and the presence of a hesitation all interact to predict the presence of a misproduction?
#

# outcome is binary
wordfreq_model_with_absents_as_median_6_z_scored_logistic <- glmer(misprod_outcome ~ log10frequency_with_absents_as_median_z * hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family="binomial")
# does not converge
summary(wordfreq_model_with_absents_as_median_6_z_scored)
summary(wordfreq_model_with_absents_as_median_6_z_scored_logistic)


# try again: raise # of iterations
Expand All @@ -864,32 +834,56 @@ wordfreq_model_with_absents_as_median_6_z_scored_logistic_no_psg_bobyqa <-
summary(wordfreq_model_with_absents_as_median_6_z_scored_logistic_no_psg_bobyqa)


# Comprehension models

# ???
# "boundary (singular) fit: see help('isSingular')" for all of the below models

# Social anxiety to predict comprehension accuracy
# n.s.
# scaaredSoc_z -0.05235 0.18485 -0.283 0.777
word_level_comprehension_model_0_logistic <-
glmer(challengeACC_outcome ~ scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(word_level_comprehension_model_0_logistic)

# Misproduction to predict comprehension accuracy
# n.s.
# misprod_predictor1 0.009751 0.023084 0.422 0.673
word_level_comprehension_model_1_logistic <-
glmer(challengeACC_outcome ~ misprod_predictor + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(word_level_comprehension_model_1_logistic)

# Hesitation to predict comprehension accuracy
# n.s.
# hesitation_predictor1 0.03030 0.02167 1.398 0.162
word_level_comprehension_model_2_logistic <-
glmer(challengeACC_outcome ~ hesitation_predictor + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(word_level_comprehension_model_2_logistic)

# Interaction between misproduction and social anxiety to predict comprehension accuracy
# n.s.
# misprod_predictor1 0.00975 0.02309 0.422 0.673
# scaaredSoc_z -0.04265 0.19825 -0.215 0.830
# misprod_predictor1:scaaredSoc_z 0.01022 0.02119 0.482 0.630
word_level_comprehension_model_3_logistic <-
glmer(challengeACC_outcome ~ misprod_predictor * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(word_level_comprehension_model_3_logistic)

# Interaction between hesitation and social anxiety to predict comprehension accuracy
# n.s.
# hesitation_predictor1 0.031587 0.022160 1.425 0.154
# scaaredSoc_z -0.058026 0.193953 -0.299 0.765
# hesitation_predictor1:scaaredSoc_z -0.005708 0.020126 -0.284 0.777
word_level_comprehension_model_4_logistic <-
glmer(challengeACC_outcome ~ hesitation_predictor * scaaredSoc_z + (1|id) + (1|passage) + (1|word),
data=errorDat, family = "binomial")
summary(word_level_comprehension_model_4_logistic)


# todo check comprehension
# word_level_comprehension_model_0 <-
# glmer(challengeACC ~ scaaredSoc + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(word_level_comprehension_model_0)
#
# word_level_comprehension_model_1 <-
# glmer(challengeACC ~ misprod + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(word_level_comprehension_model_1)
#
# word_level_comprehension_model_2 <-
# glmer(challengeACC ~ hesitation + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(word_level_comprehension_model_2)
#
# word_level_comprehension_model_3 <-
# glmer(challengeACC ~ misprod * scaaredSoc + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(word_level_comprehension_model_3)
#
# word_level_comprehension_model_4 <-
# glmer(challengeACC ~ hesitation * scaaredSoc + (1|id) + (1|passage) + (1|word),
# data=errorDat, family = "binomial")
# summary(word_level_comprehension_model_4)

# effects
# eff <- effect("log10frequency_with_absents_as_median", wordfreq_model_with_absents_as_median_1)
Expand Down

0 comments on commit 5424a12

Please sign in to comment.