1 Set up R environment
2 Read-in datafiles
3 Cognitive Styles
4 IB performance
5 Post IB questions
- 5.1 Expecting an unexpected event
- 5.2 Familiarity with IB experiments
6 Cognitive style ratio predicting IB

1 Set up R environment

library(tidyverse)
library(ggplot2)
library(ggpubr)
library(plyr)

Set the R working drectory to the main experiment directory.

setwd("/Users/adambarnas/Box/CogStyles_IB")

2 Read-in datafiles

Read in the individual subject files.

Get a count of the number of subjects.

nrow(tbl_all %>% distinct(ID,.keep_all = FALSE))

## [1] 65

3 Cognitive Styles

Organize data

tbl_all_simple = subset(tbl_all, select = -c(rowNo,responseWindow,presTime,ISI,ITI,stimFormat,button1,keyboard,key,responseType,randomPick,responseOptions,pageBreak,required,if.,then,ITI_ms,ITI_f,ITI_fDuration,presTime_ms,presTime_f,presTime_fDuration,timestamp,responseCode,correct))

tbl_all_simple <- tbl_all_simple %>% 
  separate(response,into=c('response'))

tbl_all_simple <- tbl_all_simple %>% 
  separate(RTkeys,into=c('RTkeys'))

tbl_all_simple[tbl_all_simple==""]<-NA

tbl_all_simple$rt<- tbl_all_simple$RTkeys
tbl_all_simple$rt[is.na(tbl_all_simple$rt)] <- tbl_all_simple$RT[is.na(tbl_all_simple$rt)]

tbl_all_simple$rt  <- as.numeric(tbl_all_simple$rt)
tbl_all_simple$ID  <- as.character(tbl_all_simple$ID)
tbl_all_simple$stim1  <- as.character(tbl_all_simple$stim1)

tbl_all_blank_removed <- tbl_all_simple %>% 
  filter(stim1 != "blank")

tbl_all_cog_style <- tbl_all_blank_removed %>% 
  filter(grepl('embedded|matching', stim1))

tbl_all_cog_style[tbl_all_cog_style== "embedded_prompt" ] <- NA
tbl_all_cog_style[tbl_all_cog_style== "matching_prompt" ] <- NA
tbl_all_cog_style[tbl_all_cog_style== "timeout" ] <- NA

tbl_all_cog_style <- tbl_all_cog_style %>% 
  group_by(ID) %>% 
  fill(stim1) %>% #default direction down
  fill(stim1, .direction = "up")

tbl_all_cog_style_condensed <- tbl_all_cog_style %>% 
  group_by(ID, stim1) %>% 
  dplyr::summarise(RT = sum(rt), response = first(na.omit(response)))

Split up the cognitive styles stimuli nomenclature to task, answer, and number. Filter the two cognitive styles tasks.

tbl_all_cog_style_condensed <- tbl_all_cog_style_condensed %>%
  separate(stim1,into=c('task', 'answer', 'number'))

3.1 Analyze accuracy

In the matching figures task, subjects were instructed to press ‘F’ if the two complex shapes were the same and ‘J’ if the two complex shapes were different. Trials will be labeled 1 for correct responses (‘F’ for same objects and ‘J’ for different objects) and 0 for incorrect responses (‘F’ for different objects and ‘J’ for same objects).

In the embedded figures task, subjects were instructed to press ‘F’ if the simple shape is within the complex shape and ‘J’ if the the simple shape is not within the complex shape. Trials will be labeled 1 for correct responses (‘F’ for simple within complex and ‘J’ for simple not within complex) and 0 for incorrect responses (‘F’ simple not within complex and ‘J’ simple within complex).

tbl_all_cog_style_condensed$acc = "filler"

for (i in 1:length(tbl_all_cog_style_condensed$ID)){
  if (tbl_all_cog_style_condensed$task[i] == "matching"){
    if (tbl_all_cog_style_condensed$answer[i] == "same"){
      if (tbl_all_cog_style_condensed$response[i] == "f"){
        tbl_all_cog_style_condensed$acc[i] = 1
    } else {
        tbl_all_cog_style_condensed$acc[i] = 0
    }
    } else {
      if (tbl_all_cog_style_condensed$response[i] == "j"){
        tbl_all_cog_style_condensed$acc[i] = 1
    } else {
        tbl_all_cog_style_condensed$acc[i] = 0
    }
  }
  }
  if (tbl_all_cog_style_condensed$task[i] == "embedded"){
    if (tbl_all_cog_style_condensed$answer[i] == "yes"){
      if (tbl_all_cog_style_condensed$response[i] == "f"){
        tbl_all_cog_style_condensed$acc[i] = 1
    } else {
        tbl_all_cog_style_condensed$acc[i] = 0
    }
    } else {
      if (tbl_all_cog_style_condensed$response[i] == "j"){
        tbl_all_cog_style_condensed$acc[i] = 1
    } else {
        tbl_all_cog_style_condensed$acc[i] = 0
    }
  }
  }
}

write.csv(tbl_all_cog_style_condensed,'embedded_matching_data.csv', row.names=FALSE)

3.2 Plot accuracy

tbl_all_cog_style_condensed_acc <- tbl_all_cog_style_condensed %>%
  group_by(ID,task,acc) %>%
  dplyr::summarize(counts = n()) %>%
  spread(acc,counts) %>% 
  mutate(total = rowSums(.[3:4], na.rm = TRUE))
colnames(tbl_all_cog_style_condensed_acc) <- c("ID", "task", "inacc", "acc", "total")
tbl_all_cog_style_condensed_acc[is.na(tbl_all_cog_style_condensed_acc)] <- 0
tbl_all_cog_style_condensed_acc$rate <- tbl_all_cog_style_condensed_acc$acc / tbl_all_cog_style_condensed_acc$total

tbl_all_cog_style_condensed_acc %>%
  ggbarplot("ID", "rate", fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), font.xtickslab = 6, ylab = "Accuracy", ylim = c(0, 1), position = position_dodge(0.8)) + rotate_x_text() + geom_hline(yintercept = .5, linetype = 2)

tbl_all_cog_style_condensed_acc %>%
  ggbarplot("task", "rate", add = "mean_se",fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), ylab = "Accuracy", ylim = c(0, 1), position = position_dodge(0.8)) + geom_hline(yintercept = .5, linetype = 2)

embedded_chance_sona <- tbl_all_cog_style_condensed_acc %>% 
  filter(task =="embedded")
embedded_chance_sona <-t.test(embedded_chance_sona$rate, mu = .50, alternative="greater")
embedded_chance_sona

## 
##  One Sample t-test
## 
## data:  embedded_chance_sona$rate
## t = 16.99, df = 64, p-value < 2.2e-16
## alternative hypothesis: true mean is greater than 0.5
## 95 percent confidence interval:
##  0.7871772       Inf
## sample estimates:
## mean of x 
## 0.8184615

matching_chance_sona <- tbl_all_cog_style_condensed_acc %>% 
  filter(task =="matching")
matching_chance_sona <-t.test(matching_chance_sona$rate, mu = .50, alternative="greater")
matching_chance_sona

## 
##  One Sample t-test
## 
## data:  matching_chance_sona$rate
## t = 15.229, df = 64, p-value < 2.2e-16
## alternative hypothesis: true mean is greater than 0.5
## 95 percent confidence interval:
##  0.7256833       Inf
## sample estimates:
## mean of x 
## 0.7534615

tbl_all_cog_style_condensed_acc %>% 
  with(t.test(rate~task,paired=TRUE))

## 
##  Paired t-test
## 
## data:  rate by task
## t = 4.8539, df = 64, p-value = 8.124e-06
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.0382478 0.0917522
## sample estimates:
## mean of the differences 
##                   0.065

3.3 Remove subjects below chance

tbl_all_cog_style_condensed_acc <- subset(tbl_all_cog_style_condensed_acc, select = c(ID, task, rate))
tbl_all_cog_style_condensed_acc_wide <- tbl_all_cog_style_condensed_acc %>%
  spread(task,rate)

tbl_all_cog_style_condensed_acc <- tbl_all_cog_style_condensed_acc_wide %>%
  filter(embedded > 0.5 & matching > 0.5)

tbl_all_cog_style_condensed_acc <- gather(tbl_all_cog_style_condensed_acc, task, rate, embedded:matching, factor_key=TRUE)

tbl_all_cog_style_condensed_acc %>%
  ggbarplot("ID", "rate", fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), font.xtickslab = 6, ylab = "Accuracy", ylim = c(0, 1), position = position_dodge(0.8)) + rotate_x_text() + geom_hline(yintercept = .5, linetype = 2)

tbl_all_cog_style_condensed_acc %>%
  ggbarplot("task", "rate", add = "mean_se",fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), ylab = "Accuracy", ylim = c(0, 1), position = position_dodge(0.8)) + geom_hline(yintercept = .5, linetype = 2)

embedded_chance_sona <- tbl_all_cog_style_condensed_acc %>% 
  filter(task =="embedded")
embedded_chance_sona <-t.test(embedded_chance_sona$rate, mu = .50, alternative="greater")
embedded_chance_sona

## 
##  One Sample t-test
## 
## data:  embedded_chance_sona$rate
## t = 29.051, df = 56, p-value < 2.2e-16
## alternative hypothesis: true mean is greater than 0.5
## 95 percent confidence interval:
##  0.8414234       Inf
## sample estimates:
## mean of x 
## 0.8622807

matching_chance_sona <- tbl_all_cog_style_condensed_acc %>% 
  filter(task =="matching")
matching_chance_sona <-t.test(matching_chance_sona$rate, mu = .50, alternative="greater")
matching_chance_sona

## 
##  One Sample t-test
## 
## data:  matching_chance_sona$rate
## t = 21.68, df = 56, p-value < 2.2e-16
## alternative hypothesis: true mean is greater than 0.5
## 95 percent confidence interval:
##  0.7659282       Inf
## sample estimates:
## mean of x 
## 0.7881579

tbl_all_cog_style_condensed_acc %>% 
  with(t.test(rate~task,paired=TRUE))

## 
##  Paired t-test
## 
## data:  rate by task
## t = 5.4236, df = 56, p-value = 1.289e-06
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.04674492 0.10150069
## sample estimates:
## mean of the differences 
##              0.07412281

3.4 Plot RTs

tbl_all_cog_style_condensed_rts <- tbl_all_cog_style_condensed[(tbl_all_cog_style_condensed$ID %in% tbl_all_cog_style_condensed_acc$ID),] %>% 
  filter(acc == 1)

tbl_all_cog_style_condensed_rts %>%
  ggbarplot("ID", "RT", fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), font.xtickslab = 6, add = "median", position = position_dodge(0.8), ylab = "Median RT (ms)", ylim = c(0,6000)) + rotate_x_text()

tbl_all_cog_style_condensed_rts %>%
  ggbarplot("task", "RT", add = "median",fill = "task", color = "task", palette = c("#0d2240", "#00a8e1"), position = position_dodge(0.8), order = c("embedded", "matching"), ylab = "Median RT (ms)", ylim = c(0,5000))

tbl_all_cog_style_condensed_rts_median <- tbl_all_cog_style_condensed_rts %>%
  group_by(ID,task) %>%
  dplyr::summarize(median_rt = median(RT, na.rm=TRUE))

tbl_all_cog_style_condensed_rts_median %>% 
  with(t.test(median_rt~task,paired=TRUE))

## 
##  Paired t-test
## 
## data:  median_rt by task
## t = -15.089, df = 56, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1438.777 -1101.521
## sample estimates:
## mean of the differences 
##               -1270.149

3.5 Compute cognitive style ratio

The wholist-analytic ratio is calculated by dividing the median response latency to items in the matching figures task (wholist) by the median response latency to items in the embedded figures task (analytic) A ratio of below 1 indicates that an individual responded relatively faster to the matching figure items, corresponding to a wholist profile; a ratio of above 1 indicates that an individual responded relatively faster to the embedded figure items, corresponding to a analytic profile.

tbl_all_cog_style_condensed_rts_median <- tbl_all_cog_style_condensed_rts_median %>% 
  spread(task, median_rt)
tbl_all_cog_style_condensed_rts_median$ratio <- tbl_all_cog_style_condensed_rts_median$matching / tbl_all_cog_style_condensed_rts_median$embedded

tbl_all_cog_style_condensed_rts_median$style = "filler"
for (i in 1:length(tbl_all_cog_style_condensed_rts_median$ID)){
  if (tbl_all_cog_style_condensed_rts_median$ratio[i] > 1){
    tbl_all_cog_style_condensed_rts_median$style[i] = "analytic"
  } else {
    tbl_all_cog_style_condensed_rts_median$style[i] = "wholist"
  }
}

tbl_all_cog_style_condensed_rts_median %>%
  ggbarplot("ID", "ratio", fill = "#f7a800", color = "#f7a800", ylim = c(0,4), font.xtickslab = 6, ylab = "Median wholist-analytic ratio") + rotate_x_text() + geom_hline(yintercept = 1, linetype = 2)

table(tbl_all_cog_style_condensed_rts_median$style)

## 
## analytic 
##       57

4 IB performance

tbl_all_IB <- tbl_all %>% 
  filter(grepl('Did you notice|item', head))
tbl_all_IB = subset(tbl_all_IB, select = -c(stim1,ITI,ISI,presTime,if.,then,presTime_ms,presTime_f,presTime_fDuration,RTkeys,correct,stimFormat,button1,keyboard,key,responseWindow,randomBlock,responseType,randomPick,responseOptions,pageBreak,required,ITI_ms,ITI_f,ITI_fDuration,responseCode))

tbl_all_IB <- tbl_all_IB[(tbl_all_IB$ID %in% tbl_all_cog_style_condensed_rts_median$ID),]

4.1 Counting task

tbl_all_counting <- tbl_all %>% 
  filter(rowNo >= 406 & rowNo <= 433)
tbl_all_counting = subset(tbl_all_counting, select = -c(rowNo,type,timestamp,RT,correct,ISI,ITI,stimFormat,button1,keyboard,key,responseWindow,randomBlock,presTime,head,responseType,randomPick,responseOptions,pageBreak,required,if.,then,ITI_ms,ITI_f,ITI_fDuration,presTime_ms,presTime_f,presTime_fDuration,RTkeys,responseCode))

tbl_all_counting <- tbl_all_counting[(tbl_all_counting$ID %in% tbl_all_cog_style_condensed_rts_median$ID),]

tbl_all_counting <- cbind(tbl_all_counting[c(TRUE, FALSE), ],tbl_all_counting[c(FALSE, TRUE), ])

tbl_all_counting = subset(tbl_all_counting, select = -c(3:5))

tbl_counts <- read_csv("./IB_stims/IB_counting.csv")

tbl_counts_comparison <- full_join(tbl_all_counting, tbl_counts, by = "stim1")
tbl_counts_comparison <- tbl_counts_comparison[complete.cases(tbl_counts_comparison), ]

tbl_counts_comparison$response  <- as.numeric(tbl_counts_comparison$response)
tbl_counts_comparison$count  <- as.numeric(tbl_counts_comparison$count)
tbl_counts_comparison$error <- (abs(tbl_counts_comparison$response - tbl_counts_comparison$count) / tbl_counts_comparison$count) * 100

tbl_counts_comparison %>%
  ggbarplot("ID", "error", fill = "#f7a800", color = "#f7a800", ylim = c(0,150), font.xtickslab = 6, add = "mean_se", ylab = "Mean Percent Error on Midline-crossings Task") + rotate_x_text() + geom_hline(yintercept = 50, linetype = 2)

tbl_counts_comparison_average <- tbl_counts_comparison %>% 
  group_by(ID) %>% 
  dplyr::summarize(mean_error = mean(error, na.rm=TRUE))

tbl_counts_comparison_average %>%
  ggbarplot(y = "mean_error", fill = "#f7a800", color = "#f7a800", ylim = c(0,100), sort.val = c("asc") ,xlab = "Group", add = "mean_se", ylab = "Mean Percent Error on Midline-crossings Task")

error <-t.test(tbl_counts_comparison_average$mean_error, mu = 0, alternative="greater")
error

## 
##  One Sample t-test
## 
## data:  tbl_counts_comparison_average$mean_error
## t = 5.2325, df = 56, p-value = 1.295e-06
## alternative hypothesis: true mean is greater than 0
## 95 percent confidence interval:
##  11.97183      Inf
## sample estimates:
## mean of x 
##  17.59633

write.csv(tbl_counts_comparison,'IB_counting.csv', row.names=FALSE)

4.2 Remove subjects greater than 50%

tbl_counts_comparison_average_good <- tbl_counts_comparison_average %>%
  filter(mean_error < 50)

tbl_counts_comparison_good <- tbl_counts_comparison[(tbl_counts_comparison$ID %in% tbl_counts_comparison_average_good$ID),]

tbl_counts_comparison_good %>%
  ggbarplot("ID", "error", fill = "#f7a800", color = "#f7a800", ylim = c(0,50), font.xtickslab = 6, add = "mean_se", ylab = "Mean Percent Error on Midline-crossings Task") + rotate_x_text()

tbl_counts_comparison_average_good %>%
  ggbarplot(y = "mean_error", fill = "#f7a800", color = "#f7a800", ylim = c(0,100), sort.val = c("asc") ,xlab = "Group", add = "mean_se", ylab = "Mean Percent Error on Midline-crossings Task")

error_good <-t.test(tbl_counts_comparison_average_good$mean_error, mu = 0, alternative="greater")
error_good

## 
##  One Sample t-test
## 
## data:  tbl_counts_comparison_average_good$mean_error
## t = 6.2154, df = 47, p-value = 6.347e-08
## alternative hypothesis: true mean is greater than 0
## 95 percent confidence interval:
##  5.626806      Inf
## sample estimates:
## mean of x 
##  7.707548

4.3 Noticers vs. Non-noticers

tbl_all_IB <- tbl_all_IB[(tbl_all_IB$ID %in% tbl_counts_comparison_average_good$ID),]

tbl_all_notice <- tbl_all_IB %>% 
  filter(grepl('items', head))
table(tbl_all_notice$response)

## 
##       No Not sure      Yes 
##       21       12       15

tbl_all_notice = subset(tbl_all_notice, select = -c(rowNo,type,head,timestamp,RT))

write.csv(tbl_all_notice,'IB_notice.csv', row.names=FALSE)

4.4 Unexpected event description

tbl_all_event <- tbl_all_IB %>% 
  filter(grepl('moved', head))
tbl_all_event <- cbind.data.frame(tbl_all_event[1], tbl_all_event[6])
knitr::kable(tbl_all_event)

ID	response
A1969Q0R4Y0E3J	I am guessing one of those items were present and moved at a slower rate than others.
A270HX8LH9LJ8W	I didn’t consciously see one of those items.
A171RZ3O028XF6	maybe the E or plus sign
AYFDUTHXE54I	I saw letters T L crossing a center line where as L is in white color.
A3IXWFXJC2YKU7	The large X perhaps
A2OPYRV3GLAPS1	I didn’t notice anything.
AD1WGUMVD6KED	I think there was a plus sign somewhere.
A1IQUHRI7G63YF	The plus sign appeared on the line
A2XQ3CFB5HT2ZQ	no idea
AMPMTF5IAAMK8	an x moved on a horizontal line
A2HRUFTA09371Y	#NAME?
A3LT7W355XOAKF	x, only a guess, moved the same as the others
A1SHLWKA0UH1IS	I did not notice it.
AM8OWAW9TUVLN	I just watched the white ones and forgot what they were
A1U4DNYKF6YUY3	+, moved along the center horizontal line from the right to the left.
A3RHJEMZ4EGY2U	It seemed that something was moving faster but I have no idea what it was because I was busy trying to count.
A2HJDLSU95MH8Q	no idea, sorry
A15XIKX3PI7SLK	I did not notice any of the items.
A3N0QZ9ZKUCTCQ	Plus sign, moving slowly right to left.
A2FOYHZ7HOFKBI	Didn’t see any of the items, I was too focused on the white letters
A1W7I6FN183I8F	I did not see those letters/equation appear at all while I was counting.
A1P3HHEXWNLJMP	I did not notice it at all.
A1JR35HATOEME9	They moved slower
ACKG8OU1KHKO2	plus sign moved on the left side on the line.
AD21ES5UMD77U	I did not see any of the previous items
A2OX8TSRCU6NKD	I did not see anything different but if there was my guess was it didn’t move
A1N0Q3QP4OMTZ2	It moved right on the center line, and it was the third image (grey cross).
A2I6ZALE49CVSC	It just moved across the screen
A279YSE7NB16M0	no
AONN1W54VC8YD	The + symbol moved to left, along the line.
A198MSVO1VTAT5	I didn’t notice it at all. I was watching the L T.
A2K287FPB9YFIE	I didn’t pay attention to the dark symbols.
A3APKUC67F9IMW	I’m sure I saw crosses but not in the dimensions or rotation that you have there. You got me worried when you hit me right in the gut immediately.
A22VZZFOP9D3GC	x
A1QKIA8XRNEXIG	I’m not sure, I didn’t notice any unusual items in the animation
A1O0BGHFTMPQM0	DIDN’T NOTICE IT
A3EC3OP6U52JYC	I saw a gray cross move from right to left across the middle of the screen.
AKVDK30EEV08C	I didn’t notice anything different.
A31JM9RECQGYEX	none of them were present
A297OTX4PW0XS3	The plus symbol moved along the line from one side to the other.
A2R75YFKVALBXE	Didn’t see it.
AFIK3VBMMX6G6	plus sign, right to left
AQN3WMCEA96DQ	I was only paying attention to the white figures. Maybe an X crossed the screen though. just a guess
A1Q4ZOXZ0Y7K6I	I saw a gray plus shape moving from right to left on the line
A3OP24TYKA619W	i did not see it. too busy concentrating
A2POU9TTW177VH	I did not see it
A37BGU8FSIP4I6	T L
A2ZDEERVRN5AMC	I was too focused on the purple line.

5 Post IB questions

5.1 Expecting an unexpected event

tbl_all <- tbl_all[(tbl_all$ID %in% tbl_all_IB$ID),]

tbl_all_expecting <- tbl_all %>% 
  filter(grepl('Before', head))
tbl_all_expecting = subset(tbl_all_expecting, select = -c(stim1,ITI,ISI,presTime,if.,then,presTime_ms,presTime_f,presTime_fDuration,RTkeys,correct,stimFormat,button1,keyboard,key,responseWindow,randomBlock,responseType,randomPick,responseOptions,pageBreak,required,ITI_ms,ITI_f,ITI_fDuration,responseCode))

tbl_all_expecting <- tbl_all_expecting[(tbl_all_expecting$ID %in% tbl_all_cog_style_condensed_rts_median$ID),]

table(tbl_all_expecting$response)

## 
##  No Yes 
##  45   3

5.2 Familiarity with IB experiments

tbl_all_familiarity <- tbl_all %>% 
  filter(grepl('gorilla', head))
tbl_all_familiarity = subset(tbl_all_familiarity, select = -c(stim1,ITI,ISI,presTime,if.,then,presTime_ms,presTime_f,presTime_fDuration,RTkeys,correct,stimFormat,button1,keyboard,key,responseWindow,randomBlock,responseType,randomPick,responseOptions,pageBreak,required,ITI_ms,ITI_f,ITI_fDuration,responseCode))

tbl_all_familiarity <- tbl_all_familiarity[(tbl_all_familiarity$ID %in% tbl_all_cog_style_condensed_rts_median$ID),]

table(tbl_all_familiarity$response)

## 
##  No Yes 
##  34  14

6 Cognitive style ratio predicting IB

tbl_log_reg_median <- merge(tbl_all_cog_style_condensed_rts_median, tbl_all_notice, by = "ID")
#tbl_log_reg_median = subset(tbl_log_reg_median, select = -c(rowNo,type,head,timestamp,RT))
tbl_log_reg_median[tbl_log_reg_median == "Yes"] <- 1
tbl_log_reg_median[tbl_log_reg_median == "No"] <- 0
tbl_log_reg_median[tbl_log_reg_median == "Not sure"] <- 0
names(tbl_log_reg_median)[names(tbl_log_reg_median)=="response"] <- "notice"
tbl_log_reg_median$notice <- as.numeric(tbl_log_reg_median$notice)

6.1 Median RTs

6.1.1 Logistic regression w/ median RT ratio

log_reg_median_ratio <- glm(notice ~ ratio, data = tbl_log_reg_median, family = binomial(link = "logit"))
summary(log_reg_median_ratio)

## 
## Call:
## glm(formula = notice ~ ratio, family = binomial(link = "logit"), 
##     data = tbl_log_reg_median)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.9599  -0.8779  -0.8372   1.4568   1.7207  
## 
## Coefficients:
##             Estimate Std. Error z value Pr(>|z|)
## (Intercept)  -0.1429     1.3656  -0.105    0.917
## ratio        -0.3461     0.7180  -0.482    0.630
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 59.386  on 46  degrees of freedom
## AIC: 63.386
## 
## Number of Fisher Scoring iterations: 4

plot_median_ratio <- ggplot(tbl_log_reg_median, aes(x=ratio, y=notice)) + xlim(1,4) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_median_ratio))

6.1.2 Logistic regression w/ embedded median RT

log_reg_embedded_median <- glm(notice ~ embedded, data = tbl_log_reg_median, family = binomial(link = "logit"))
summary(log_reg_embedded_median)

## 
## Call:
## glm(formula = notice ~ embedded, family = binomial(link = "logit"), 
##     data = tbl_log_reg_median)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.9790  -0.8878  -0.8354   1.4098   1.8595  
## 
## Coefficients:
##               Estimate Std. Error z value Pr(>|z|)
## (Intercept)  0.4501601  1.5725125   0.286    0.775
## embedded    -0.0007989  0.0010054  -0.795    0.427
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 58.956  on 46  degrees of freedom
## AIC: 62.956
## 
## Number of Fisher Scoring iterations: 4

plot_embedded_median <- ggplot(tbl_log_reg_median, aes(x=embedded, y=notice)) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_embedded_median))

6.1.3 Logistic regression w/ matching median RT

log_reg_matching_median <- glm(notice ~ matching, data = tbl_log_reg_median, family = binomial(link = "logit"))
summary(log_reg_matching_median)

## 
## Call:
## glm(formula = notice ~ matching, family = binomial(link = "logit"), 
##     data = tbl_log_reg_median)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -1.0319  -0.8921  -0.7818   1.3609   1.9156  
## 
## Coefficients:
##               Estimate Std. Error z value Pr(>|z|)
## (Intercept)  0.5517346  1.3364426   0.413    0.680
## matching    -0.0004711  0.0004647  -1.014    0.311
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 58.519  on 46  degrees of freedom
## AIC: 62.519
## 
## Number of Fisher Scoring iterations: 4

plot_matching_median <- ggplot(tbl_log_reg_median, aes(x=matching, y=notice)) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_matching_median))

6.1.4 Logistic regression (Dichotomous predictor, dichotomous outcome)

Does not work because there are no ‘wholists’

#log_reg_style <- glm(notice ~ style, data = tbl_log_reg, family = binomial)
#summary(log_reg_style)

6.2 Accuracy

tbl_all_cog_style_condensed_acc_wide <- tbl_all_cog_style_condensed_acc_wide[(tbl_all_cog_style_condensed_acc_wide$ID %in% tbl_log_reg_median$ID),]

tbl_log_reg_acc <- merge(tbl_all_cog_style_condensed_acc_wide, tbl_all_notice, by = "ID")
#tbl_log_reg_acc = subset(tbl_log_reg_acc, select = -c(rowNo,type,head,timestamp,RT))
tbl_log_reg_acc[tbl_log_reg_acc == "Yes"] <- 1
tbl_log_reg_acc[tbl_log_reg_acc == "No"] <- 0
tbl_log_reg_acc[tbl_log_reg_acc == "Not sure"] <- 0
names(tbl_log_reg_acc)[names(tbl_log_reg_acc)=="response"] <- "notice"
tbl_log_reg_acc$notice <- as.numeric(tbl_log_reg_acc$notice)
tbl_log_reg_acc$ratio <- tbl_log_reg_acc$matching / tbl_log_reg_acc$embedded

6.2.1 Logistic regression w/ accuracy ratio

log_reg_acc_ratio <- glm(notice ~ ratio, data = tbl_log_reg_acc, family = binomial(link = "logit"))
summary(log_reg_acc_ratio)

## 
## Call:
## glm(formula = notice ~ ratio, family = binomial(link = "logit"), 
##     data = tbl_log_reg_acc)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.9427  -0.8701  -0.8542   1.4859   1.5967  
## 
## Coefficients:
##             Estimate Std. Error z value Pr(>|z|)
## (Intercept)  -0.1381     2.5579  -0.054    0.957
## ratio        -0.7011     2.7418  -0.256    0.798
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 59.559  on 46  degrees of freedom
## AIC: 63.559
## 
## Number of Fisher Scoring iterations: 4

plot_acc_ratio <- ggplot(tbl_log_reg_acc, aes(x=ratio, y=notice)) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_acc_ratio))

6.2.2 Logistic regression w/ embedded accuracy

log_reg_embedded_acc <- glm(notice ~ embedded, data = tbl_log_reg_acc, family = binomial(link = "logit"))
summary(log_reg_embedded_acc)

## 
## Call:
## glm(formula = notice ~ embedded, family = binomial(link = "logit"), 
##     data = tbl_log_reg_acc)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.9341  -0.8696  -0.8385   1.5007   1.5694  
## 
## Coefficients:
##             Estimate Std. Error z value Pr(>|z|)
## (Intercept)  -0.0378     2.7458  -0.014    0.989
## embedded     -0.8703     3.1683  -0.275    0.784
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 59.549  on 46  degrees of freedom
## AIC: 63.549
## 
## Number of Fisher Scoring iterations: 4

plot_embedded_acc <- ggplot(tbl_log_reg_acc, aes(x=embedded, y=notice)) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_embedded_acc))

6.2.3 Logistic regression w/ matching accuracy

log_reg_matching_acc <- glm(notice ~ matching, data = tbl_log_reg_acc, family = binomial(link = "logit"))
summary(log_reg_matching_acc)

## 
## Call:
## glm(formula = notice ~ matching, family = binomial(link = "logit"), 
##     data = tbl_log_reg_acc)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -1.0515  -0.8823  -0.7937   1.4271   1.6460  
## 
## Coefficients:
##             Estimate Std. Error z value Pr(>|z|)
## (Intercept)    1.202      2.596   0.463    0.643
## matching      -2.509      3.265  -0.768    0.442
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 59.624  on 47  degrees of freedom
## Residual deviance: 59.032  on 46  degrees of freedom
## AIC: 63.032
## 
## Number of Fisher Scoring iterations: 4

plot_matching_acc <- ggplot(tbl_log_reg_acc, aes(x=matching, y=notice)) + geom_point() + stat_smooth(method="glm", method.args=list(family="binomial"), se=TRUE, color="#f7a800") + theme_classic((base_size = 15))
suppressMessages(print(plot_matching_acc))

6.2.4 Logistic regression (Dichotomous predictor, dichotomous outcome)