Analysis E2 Verbal Suppression

Matt

5/14/2019

Load Libraries

library(papaja)
library(TalkTyping)
library(dplyr)
library(Crump)
library(xtable)
library(ggplot2)
library(ggpubr)
library(forcats)

E2 IKSI Analysis

#load E2 data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data <- E2_data %>%
           filter(letter_accuracy == 1,
                           iksis < 5000,
                           LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
                      prop_removed = modified_recursive_moving(iksis)$prop_removed)
                  
E2_aov_out <- aov(mean_iksi ~ suppression*LetterType + 
                    Error(subject/(suppression*LetterType)), E2_data)

knitr::kable(xtable(summary(E2_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	2869695.8	204978.27	NA	NA
suppression	5	1104110.3	220822.06	16.45400	0.0000000
Residuals	70	939439.9	13420.57	NA	NA
LetterType	1	3749036.6	3749036.64	21.27414	0.0004029
Residuals	14	2467151.0	176225.07	NA	NA
suppression:LetterType	5	949883.1	189976.62	13.96436	0.0000000
Residuals	70	952307.4	13604.39	NA	NA

E2_apa_print <- apa_print(E2_aov_out)
E2_means <- model.tables(E2_aov_out,"means")

First Letter one-way with linear contrasts

#load E1B data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data$suppression <- fct_relevel(E2_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_data <- E2_data %>%
           filter(letter_accuracy == 1,
                           iksis < 5000,
                           LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
                      prop_removed = modified_recursive_moving(iksis)$prop_removed)

E2_FL_data <- E2_data %>%
                filter(LetterType == "First")

E2_FL_aov_out <- aov(mean_iksi ~ suppression + 
                    Error(subject/(suppression)), E2_FL_data)

knitr::kable(xtable(summary(E2_FL_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	5257407	375529.07	NA	NA
suppression	5	2049979	409995.82	15.29138	0
Residuals	70	1876855	26812.22	NA	NA

E2_FL_apa_print <- apa_print(E2_FL_aov_out)
E2_FL_means <- model.tables(E2_FL_aov_out,"means")

Plot

E2_FL_iksi_table <- E2_FL_data %>%
              group_by(suppression) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))

E2_FL_graph_iksi <- ggplot(E2_FL_iksi_table, aes(x=suppression,
                                             y=mIKSI))+
  geom_bar(stat="identity", position="dodge", fill="grey")+
  geom_errorbar(aes(ymin=mIKSI-sem,
                    ymax=mIKSI+sem), width=.1,
                linetype="solid", position=position_dodge(.9))+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=9)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  coord_cartesian(ylim=c(100,800))+
  scale_y_continuous(minor_breaks=seq(100,800,25),
                     breaks=seq(100,800,100))+
   #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean IKSI (ms)")+
  theme(axis.title.x = element_blank())+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))
 # facet_wrap(~suppression, scales = "free_x",
 #            strip.position="bottom")

knitr::kable(E2_FL_iksi_table)

suppression	mIKSI	sem
Normal	275.1379	35.08749
SayThe	322.3288	50.24586
TueThur	348.5493	49.93300
Alphabet	455.8985	86.99283
RandLetter	636.1838	87.95005
Count	661.6534	111.44788

E2_FL_graph_iksi

contrasts

# Normal vs. All
E2_FL_NvsAll <- apa_print(t_contrast_rm(df=E2_FL_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("SayThe","TueThur","Alphabet","RandLetter","Count"),
                             contrast_weights = c(-1,1/5,1/5,1/5,1/5,1/5) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 4.4699, df = 14, p-value = 0.0005288
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  109.1250 310.4447
## sample estimates:
## mean of x 
##  209.7848

# Normal vs. the and tuesday thursday
E2_FL_NvsTheThu <- apa_print(t_contrast_rm(df=E2_FL_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("SayThe","TueThur"),
                             contrast_weights = c(-1,1/2,1/2) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 2.0697, df = 14, p-value = 0.05746
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   -2.187929 122.790184
## sample estimates:
## mean of x 
##  60.30113

#  the and tuesday thursday vs. alphabet
E2_FL_AlphavsTheThu <- apa_print(t_contrast_rm(df=E2_FL_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("SayThe","TueThur"),
                             B_levels = c("Alphabet"),
                             contrast_weights = c(-1/2,-1/2,1) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 2.6434, df = 14, p-value = 0.01928
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   22.71998 218.19895
## sample estimates:
## mean of x 
##  120.4595

E2_FL_AlphavsRandCnt <- apa_print(t_contrast_rm(df=E2_FL_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("Alphabet"),
                             B_levels = c("RandLetter","Count"),
                             contrast_weights = c(-1,1/2,1/2) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 3.5561, df = 14, p-value = 0.003163
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   76.60338 309.43679
## sample estimates:
## mean of x 
##  193.0201

Middle Letter one-way with linear contrasts

#load E1B data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data$suppression <- fct_relevel(E2_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_data <- E2_data %>%
           filter(letter_accuracy == 1,
                           iksis < 5000,
                           LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
                      prop_removed = modified_recursive_moving(iksis)$prop_removed)

E2_ML_data <- E2_data %>%
                filter(LetterType == "Middle")

E2_ML_aov_out <- aov(mean_iksi ~ suppression + 
                    Error(subject/(suppression)), E2_ML_data)

knitr::kable(xtable(summary(E2_ML_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	79439.849	5674.2750	NA	NA
suppression	5	4014.308	802.8616	3.773794	0.0043972
Residuals	70	14892.258	212.7465	NA	NA

E2_ML_apa_print <- apa_print(E2_ML_aov_out)
E2_ML_means <- model.tables(E2_ML_aov_out,"means")

Plot

E2_ML_iksi_table <- E2_ML_data %>%
              group_by(suppression) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))

E2_ML_graph_iksi <- ggplot(E2_ML_iksi_table, aes(x=suppression,
                                             y=mIKSI))+
  geom_bar(stat="identity", position="dodge", fill="grey")+
  geom_errorbar(aes(ymin=mIKSI-sem,
                    ymax=mIKSI+sem), width=.1,
                linetype="solid", position=position_dodge(.9))+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=9)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  coord_cartesian(ylim=c(100,200))+
  scale_y_continuous(minor_breaks=seq(100,200,25),
                     breaks=seq(100,200,25))+
   #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean IKSI (ms)")+
  theme(axis.title.x = element_blank())+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))
 # facet_wrap(~suppression, scales = "free_x",
 #            strip.position="bottom")

knitr::kable(E2_ML_iksi_table)

suppression	mIKSI	sem
Normal	155.5772	7.318144
SayThe	154.1945	7.236293
TueThur	159.9988	8.459759
Alphabet	159.2873	7.949713
RandLetter	174.2472	12.148583
Count	164.6184	7.805628

E2_ML_graph_iksi

contrasts

# Normal vs. all but random

E2_ML_NvsAllbR <- apa_print(t_contrast_rm(df=E2_ML_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("SayThe","TueThur","Alphabet","Count"),
                             contrast_weights = c(-1,1/4,1/4,1/4,1/4) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 1.8359, df = 14, p-value = 0.08771
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.6642766  8.5593328
## sample estimates:
## mean of x 
##  3.947528

# Rand vs. Normal
E2_ML_RandvsNormal <- apa_print(t_contrast_rm(df=E2_ML_data,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("RandLetter"),
                             contrast_weights = c(-1,1) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 2.7359, df = 14, p-value = 0.01608
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   4.033847 33.306027
## sample estimates:
## mean of x 
##  18.66994

First Letter Accuracy one-way with linear contrasts

#load E1B data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data$suppression <- fct_relevel(E2_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_data <- E2_data %>%
           filter(LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_acc = mean(letter_accuracy))

E2acc_FL_data <- E2_data %>%
                filter(LetterType == "First")

E2acc_FL_aov_out <- aov(mean_acc ~ suppression + 
                    Error(subject/(suppression)), E2acc_FL_data)

knitr::kable(xtable(summary(E2acc_FL_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.1360795	0.0097200	NA	NA
suppression	5	0.0256047	0.0051209	3.001159	0.0163903
Residuals	70	0.1194424	0.0017063	NA	NA

E2acc_FL_apa_print <- apa_print(E2acc_FL_aov_out)
E2acc_FL_means <- model.tables(E2acc_FL_aov_out,"means")

Plot

E2acc_FL_table <- E2acc_FL_data %>%
              group_by(suppression) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc)))

E2acc_FL_graph_acc <- ggplot(E2acc_FL_table, aes(x=suppression,
                                             y=mAcc))+
  geom_bar(stat="identity", position="dodge", fill="grey")+
  geom_errorbar(aes(ymin=mAcc-sem,
                    ymax=mAcc+sem), width=.1,
                linetype="solid", position=position_dodge(.9))+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=9)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  coord_cartesian(ylim=c(0.7,1))+
  scale_y_continuous(minor_breaks=seq(0.7,1,.1),
                     breaks=seq(0.7,1,.1))+
   #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean Accuracy")+
  theme(axis.title.x = element_blank())+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))
 # facet_wrap(~suppression, scales = "free_x",
 #            strip.position="bottom")

knitr::kable(E2acc_FL_table)

suppression	mAcc	sem
Normal	0.9564783	0.0083601
SayThe	0.9421110	0.0093006
TueThur	0.9448931	0.0097569
Alphabet	0.9275567	0.0121435
RandLetter	0.9136387	0.0128113
Count	0.9100080	0.0255653

E2acc_FL_graph_acc

contrasts

# Normal vs. the tue/thur

E2acc_FL_NvsTheThur <- apa_print(t_contrast_rm(df=E2acc_FL_data,
                             subject = "subject",
                             dv = "mean_acc",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("SayThe","TueThur"),
                             contrast_weights = c(-1,1/2,1/2) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -1.7432, df = 14, p-value = 0.1032
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.028941753  0.002989327
## sample estimates:
##   mean of x 
## -0.01297621

#  Normal the tue/thur vs alpha rand count
E2acc_FL_NTheThvsARC <- apa_print(t_contrast_rm(df=E2acc_FL_data,
                             subject = "subject",
                             dv = "mean_acc",
                             condition = "suppression",
                             A_levels = c("Normal","SayThe","TueThur"),
                             B_levels = c("Alphabet","RandLetter","Count"),
                             contrast_weights = c(-1/3,-1/3,-1/3,1/3,1/3,1/3) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -3.0697, df = 14, p-value = 0.008319
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.052251532 -0.009267813
## sample estimates:
##   mean of x 
## -0.03075967

Middle Letter Accuracy one-way with linear contrasts

#load E1B data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data$suppression <- fct_relevel(E2_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_data <- E2_data %>%
           filter(LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_acc = mean(letter_accuracy))

E2acc_ML_data <- E2_data %>%
                filter(LetterType == "Middle")

E2acc_ML_aov_out <- aov(mean_acc ~ suppression + 
                    Error(subject/(suppression)), E2acc_ML_data)

knitr::kable(xtable(summary(E2acc_ML_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.2990676	0.0213620	NA	NA
suppression	5	0.0801541	0.0160308	5.596677	0.0002145
Residuals	70	0.2005042	0.0028643	NA	NA

E2acc_ML_apa_print <- apa_print(E2acc_ML_aov_out)
E2acc_ML_means <- model.tables(E2acc_ML_aov_out,"means")

Plot

E2acc_ML_table <- E2acc_ML_data %>%
              group_by(suppression) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc)))

E2acc_ML_graph_acc <- ggplot(E2acc_ML_table, aes(x=suppression,
                                             y=mAcc))+
  geom_bar(stat="identity", position="dodge", fill="grey")+
  geom_errorbar(aes(ymin=mAcc-sem,
                    ymax=mAcc+sem), width=.1,
                linetype="solid", position=position_dodge(.9))+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=9)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  coord_cartesian(ylim=c(0.7,1))+
  scale_y_continuous(minor_breaks=seq(0.7,1,.1),
                     breaks=seq(0.7,1,.1))+
   #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean Accuracy")+
  theme(axis.title.x = element_blank()) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))
 # facet_wrap(~suppression, scales = "free_x",
 #            strip.position="bottom")

knitr::kable(E2acc_ML_table)

suppression	mAcc	sem
Normal	0.8978740	0.0178426
SayThe	0.8603967	0.0176930
TueThur	0.8576019	0.0159934
Alphabet	0.8267572	0.0168839
RandLetter	0.8155491	0.0232558
Count	0.8141323	0.0258036

E2acc_ML_graph_acc

contrasts

# Normal vs. the tue/thur

E2acc_ML_NvsTheThur <- apa_print(t_contrast_rm(df=E2acc_ML_data,
                             subject = "subject",
                             dv = "mean_acc",
                             condition = "suppression",
                             A_levels = c("Normal"),
                             B_levels = c("SayThe","TueThur"),
                             contrast_weights = c(-1,1/2,1/2) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -1.9116, df = 14, p-value = 0.07662
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.082492462  0.004743188
## sample estimates:
##   mean of x 
## -0.03887464

#  Normal the tue/thur vs alpha rand count
E2acc_ML_NTheThvsARC <- apa_print(t_contrast_rm(df=E2acc_ML_data,
                             subject = "subject",
                             dv = "mean_acc",
                             condition = "suppression",
                             A_levels = c("Normal","SayThe","TueThur"),
                             B_levels = c("Alphabet","RandLetter","Count"),
                             contrast_weights = c(-1/3,-1/3,-1/3,1/3,1/3,1/3) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -4.2358, df = 14, p-value = 0.0008306
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.08005405 -0.02623533
## sample estimates:
##   mean of x 
## -0.05314469

speed accuracy tradeoff

library(ggrepel)
E2_iksi_both <- rbind(E2_FL_iksi_table,
                       E2_ML_iksi_table)

E2_iksi_both <- cbind(E2_iksi_both, 
                       Letter_Position = rep(c("First","Middle"), each=6))

E2_acc_both <- rbind(E2acc_FL_table,
                       E2acc_ML_table)

E2_acc_both <- cbind(E2_acc_both, 
                       Letter_Position = rep(c("First","Middle"), each=6))

E2_SA <- cbind(E2_iksi_both, accuracy = E2_acc_both$mAcc)

E2_SA_graph <- ggplot(E2_SA, aes(x=mIKSI,  y=accuracy,
                   shape=Letter_Position,
                   color=Letter_Position,
                   label=suppression))+
  geom_point()+
  geom_text_repel(size=1.7, color="black")+
  coord_cartesian(xlim=c(100,700), ylim=c(.8,1))+
  scale_x_continuous(breaks=seq(100,700,100))+
  scale_color_grey(start = 0.2, end = 0.5, na.value = "red",
  aesthetics = "color", guide =FALSE)+
  theme_classic(base_size=10)+
  ylab("Mean Accuracy")+
  xlab("Mean IKSI (ms)")+
  #facet_wrap(~Letter_Position, nrow=2,
  #           strip.position="right")+
  theme(legend.position ="top",
        legend.direction = "horizontal")+
  guides(shape = guide_legend(label.hjust = 0,
                              keywidth=0.1))+
  labs(shape="Letter")
  

E2_SA_graph

#load E2 data
E2_data <- talk_type_E2_data

# IKSI analysis

E2_data <- E2_data %>%
           filter(accuracy == 1,
                           iksis < 5000,
                           LetterType != "Space") %>%
            mutate(subject = as.factor(subject),
                   suppression = as.factor(suppression),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,suppression,LetterType) %>%
            summarise(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
                      prop_removed = modified_recursive_moving(iksis)$prop_removed)

E2_data$suppression <- fct_relevel(E2_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_iksi_table <- E2_data %>%
              group_by(suppression,LetterType) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))

#levels(E1A_iksi_table$paragraph) <- c("Say Letter", "Say Word")

E2_graph_iksi <- ggplot(E2_iksi_table, aes(x=LetterType,
                                             y=mIKSI,
                                           fill=LetterType))+
  geom_bar(stat="identity",position="dodge")+
  geom_errorbar(aes(ymin=mIKSI-sem,
                    ymax=mIKSI+sem), width=.1,
                linetype="solid")+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=12)+
  theme(legend.position = "none",
        legend.title = element_blank())+
  ylab("Mean IKSI (ms)")+
  xlab("Letter position")+
  facet_wrap(~suppression, nrow=1)

knitr::kable(E2_iksi_table)

suppression	LetterType	mIKSI	sem
Normal	First	276.9624	34.722182
Normal	Middle	154.1825	7.340545
SayThe	First	321.7087	53.218289
SayThe	Middle	153.2280	7.335822
TueThur	First	345.5774	48.242259
TueThur	Middle	158.1935	8.228906
Alphabet	First	477.5381	95.189790
Alphabet	Middle	158.0736	7.993724
RandLetter	First	650.1842	89.511870
RandLetter	Middle	170.7278	11.030549
Count	First	681.3246	115.473813
Count	Middle	159.9962	7.404675

E2_graph_iksi

accuracy

get letter level accuracy estimates

This chunk was previously evaluated, and the talk_type_e2_data.RData file was updated with estimates of letter level accuracy.

E2acc_data <- talk_type_E2_data

E2acc_data$closest[is.na(E2acc_data$closest)] <- ""

letter_accuracy <- E2acc_data$accuracy
for(i in 1:dim(E2acc_data)[1]){
  clet <- as.character(E2acc_data[i,]$letters)
  cword <- unlist(strsplit(as.character(E2acc_data[i,]$closest),split=""))
  if(E2acc_data$accuracy[i] == 0){
    if(length(cword)>0){
      if(E2acc_data[i,]$LetterPosition <= length(cword)){
        if(clet != " "){
          if(clet == cword[E2acc_data[i,]$LetterPosition]){
            letter_accuracy[i] <- 1
          } else {
            letter_accuracy[i] <- 0
          }
        }
      }
    }
  }
}

E2acc_data <- cbind(E2acc_data,letter_accuracy)

Accuracy ANOVA

E2acc_data <- talk_type_E2_data

# Accuracy

E2acc_data <- E2acc_data %>%
                mutate(subject = as.factor(subject),
                       suppression = as.factor(suppression),
                       LetterType = as.factor(LetterType)) %>%
                filter(LetterType != "Space") %>%
                group_by(subject,suppression, LetterType) %>%
                summarise(mean_acc = mean(letter_accuracy))
                  
E2acc_aov_out <- aov(mean_acc ~ suppression*LetterType + Error(subject/(suppression*LetterType)), E2acc_data)

E2acc_apa_print <- apa_print(E2acc_aov_out)
E2acc_means <- model.tables(E2acc_aov_out,"means")

knitr::kable(xtable(summary(E2acc_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.3972025	0.0283716	NA	NA
suppression	5	0.0965600	0.0193120	4.961481	0.0006037
Residuals	70	0.2724671	0.0038924	NA	NA
LetterType	1	0.3410941	0.3410941	125.849925	0.0000000
Residuals	14	0.0379445	0.0027103	NA	NA
suppression:LetterType	5	0.0091988	0.0018398	2.712384	0.0268243
Residuals	70	0.0474795	0.0006783	NA	NA

Accuracy table and plots

E2acc_data$suppression <- fct_relevel(E2acc_data$suppression,c("Normal","SayThe","TueThur","Alphabet","RandLetter","Count"))

E2_acc_table <- E2acc_data %>%
              group_by(suppression,LetterType) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc)))

#levels(E1A_acc_table$paragraph) <- c("Say Letter", "Say Word")

E2_graph_acc <- ggplot(E2_acc_table, aes(x=LetterType,
                                             y=mAcc,
                                           fill=LetterType))+
  geom_bar(stat="identity",position="dodge")+
  geom_errorbar(aes(ymin=mAcc-sem,
                    ymax=mAcc+sem), width=.1,
                linetype="solid")+
  scale_fill_grey(start = 0.6, end = 0.8, na.value = "red",
  aesthetics = "fill")+
  theme_classic(base_size=12)+
  theme(legend.position = "none",
        legend.title = element_blank())+
  ylab("Mean Accuracy")+
  xlab("Letter position")+
  coord_cartesian(ylim=c(.7,1))+
  facet_wrap(~suppression, nrow=1)

knitr::kable(E2_acc_table)

suppression	LetterType	mAcc	sem
Normal	First	0.9564783	0.0083601
Normal	Middle	0.8978740	0.0178426
SayThe	First	0.9421110	0.0093006
SayThe	Middle	0.8603967	0.0176930
TueThur	First	0.9448931	0.0097569
TueThur	Middle	0.8576019	0.0159934
Alphabet	First	0.9275567	0.0121435
Alphabet	Middle	0.8267572	0.0168839
RandLetter	First	0.9136387	0.0128113
RandLetter	Middle	0.8155491	0.0232558
Count	First	0.9100080	0.0255653
Count	Middle	0.8141323	0.0258036

E2_graph_acc

Figure 1

ggarrange(E2_graph_iksi,E2_graph_acc, nrow=2)

ggarrange(E2_FL_graph_iksi,E2_ML_graph_iksi, nrow=2)

Speed accuracy tradeoff

tradeoff <- E2_iksi_table
tradeoff <- cbind(tradeoff,accuracy = E2_acc_table$mAcc,
                  accuracy_sem = E2_acc_table$sem)

ggplot(tradeoff, aes(x=mIKSI, y=accuracy, 
                     group = suppression,
                     shape= LetterType,
                     linetype = suppression))+
  geom_point()+
  geom_line() +
  geom_errorbar(aes(ymin=accuracy-accuracy_sem,
                    ymax=accuracy+accuracy_sem), width=.1,
                linetype="solid")+
  geom_errorbarh(aes(xmin=mIKSI-sem,
                    xmax=mIKSI+sem), width=.2,
                linetype="solid")+
  theme_classic(base_size=12)+
  xlab("Mean IKSI (ms)")+
  ylab("Mean Accuracy")

Save all

save.image(file="E2_workspace.RData")

talk_type_E1B_data <- talk_type_E1B_data %>%
                        rename(voice = VoiceCondition,
                               linguistic_unit = ChunkCondition)

save(talk_type_E1B_data, file="talk_type_E1B_data.RData")