Analysis E1B Talk and Type

Matt

5/14/2019

Load Libraries

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

E1B IKSI Analysis

#load E1B data
E1B_data <- talk_type_E1B_data

# IKSI analysis

E1B_data <- E1B_data %>%
           filter(accuracy == 1,
                           iksis < 5000,
                           LetterType != "Space",
                           voice != "Normal",
                           linguistic_unit != "Normal") %>%
            mutate(subject = as.factor(subject),
                   voice = as.factor(voice),
                   linguistic_unit = as.factor(linguistic_unit),
                   LetterType = as.factor(LetterType)) %>%
            group_by(subject,voice,linguistic_unit,LetterType) %>%
            summarise(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
                      prop_removed = modified_recursive_moving(iksis)$prop_removed)
                  
E1B_aov_out <- aov(mean_iksi ~ voice*linguistic_unit*LetterType + 
                    Error(subject/(voice*linguistic_unit*LetterType)), E1B_data)

knitr::kable(xtable(summary(E1B_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	510011.751	36429.411	NA	NA
voice	1	12916.437	12916.437	3.842151	0.0701924
Residuals1	14	47064.807	3361.772	NA	NA
linguistic_unit	1	478311.239	478311.239	24.453078	0.0002154
Residuals	14	273845.172	19560.369	NA	NA
LetterType	1	687226.213	687226.213	24.022714	0.0002337
Residuals	14	400502.909	28607.351	NA	NA
voice:linguistic_unit	1	14780.113	14780.113	3.076148	0.1013027
Residuals	14	67266.450	4804.746	NA	NA
voice:LetterType	1	7204.638	7204.638	4.664122	0.0486337
Residuals	14	21625.708	1544.693	NA	NA
linguistic_unit:LetterType	1	22244.101	22244.101	7.815177	0.0143047
Residuals	14	39847.773	2846.270	NA	NA
voice:linguistic_unit:LetterType	1	4632.377	4632.377	3.283505	0.0914778
Residuals	14	19751.235	1410.803	NA	NA

E1B_apa_print <- apa_print(E1B_aov_out)
E1B_means <- model.tables(E1B_aov_out,"means")

E1B_iksi_table <- E1B_data %>%
              group_by(voice,linguistic_unit,LetterType) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))

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

E1B_graph_iksi <- ggplot(E1B_iksi_table, aes(x=linguistic_unit,
                                             y=mIKSI,
                                             group=LetterType,
                                             fill=LetterType))+
  geom_bar(stat="identity", position="dodge")+
  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=12)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  ylab("Mean IKSI (ms)")+
  xlab("Linguistic Unit")+
  facet_wrap(~voice)

knitr::kable(E1B_iksi_table)

voice	linguistic_unit	LetterType	mIKSI	sem
Talk	Letter	First	467.5775	38.244372
Talk	Letter	Middle	261.0719	20.618758
Talk	Word	First	279.4567	41.881543
Talk	Word	Middle	152.2636	10.242864
Silent	Letter	First	396.7084	30.105025
Silent	Letter	Middle	246.0493	18.117573
Silent	Word	First	277.8326	38.688399
Silent	Word	Middle	156.7808	9.640095

E1B_graph_iksi

accuracy

get letter level accuracy

letter level accuracies were added to the talk_type_E1B_data file.

E1Bacc_data <- talk_type_E1B_data

E1Bacc_data$closest[is.na(E1Bacc_data$closest)] <- ""

letter_accuracy <- E1Bacc_data$accuracy
for(i in 1:dim(E1Bacc_data)[1]){
  clet <- as.character(E1Bacc_data[i,]$letters)
  cword <- unlist(strsplit(as.character(E1Bacc_data[i,]$closest),split=""))
  if(E1Bacc_data$accuracy[i] == 0){
    if(length(cword)>0){
      if(E1Bacc_data[i,]$LetterPosition <= length(cword)){
        if(clet != " "){
          if(clet == cword[E1Bacc_data[i,]$LetterPosition]){
            letter_accuracy[i] <- 1
          } else {
            letter_accuracy[i] <- 0
          }
        }
      }
    }
  }
}

E1Bacc_data <- cbind(E1Bacc_data,letter_accuracy)

talk_type_E1B_data <- E1Bacc_data

# Accuracy

E1Bacc_data <- talk_type_E1B_data


E1Bacc_data <- E1Bacc_data %>%
                filter(iksis < 5000,
                       LetterType != "Space",
                       voice != "Normal",
                       linguistic_unit != "Normal") %>%
                mutate(subject = as.factor(subject),
                       voice = as.factor(voice),
                       linguistic_unit = as.factor(linguistic_unit),
                       LetterType = as.factor(LetterType)) %>%
                group_by(subject,voice,linguistic_unit, LetterType) %>%
                summarise(mean_acc = mean(letter_accuracy))
                  
E1Bacc_aov_out <- aov(mean_acc ~ voice*linguistic_unit*LetterType + Error(subject/(voice*linguistic_unit*LetterType)), E1Bacc_data)

E1Bacc_apa_print <- apa_print(E1Bacc_aov_out)
E1Bacc_means <- model.tables(E1Bacc_aov_out,"means")

knitr::kable(xtable(summary(E1Bacc_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.1631269	0.0116519	NA	NA
voice	1	0.0016173	0.0016173	2.3275604	0.1493709
Residuals1	14	0.0097281	0.0006949	NA	NA
linguistic_unit	1	0.0230912	0.0230912	11.7181925	0.0041185
Residuals	14	0.0275876	0.0019705	NA	NA
LetterType	1	0.0835370	0.0835370	31.2027809	0.0000671
Residuals	14	0.0374812	0.0026772	NA	NA
voice:linguistic_unit	1	0.0086392	0.0086392	5.1177370	0.0401165
Residuals	14	0.0236331	0.0016881	NA	NA
voice:LetterType	1	0.0020324	0.0020324	4.6876499	0.0481426
Residuals	14	0.0060700	0.0004336	NA	NA
linguistic_unit:LetterType	1	0.0029127	0.0029127	5.8468638	0.0298204
Residuals	14	0.0069742	0.0004982	NA	NA
voice:linguistic_unit:LetterType	1	0.0001564	0.0001564	0.3220731	0.5793520
Residuals	14	0.0067979	0.0004856	NA	NA

E1B_acc_table <- E1Bacc_data %>%
              group_by(voice,linguistic_unit, LetterType) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc)))

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

E1B_graph_acc <- ggplot(E1B_acc_table, 
                         aes(x=linguistic_unit,
                             y=mAcc, 
                             group=LetterType,
                             fill=LetterType))+
  geom_bar(stat="identity",position="dodge")+
  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=12)+
  theme(legend.position = "top",
        legend.title = element_blank())+
  ylab("Mean Word Accuracy")+
  xlab("Linguistic Unit")+
  coord_cartesian(ylim=c(.8,1))+
  facet_wrap(~voice)

knitr::kable(E1B_acc_table)

voice	linguistic_unit	LetterType	mAcc	sem
Talk	Letter	First	0.9829305	0.0058115
Talk	Letter	Middle	0.9340671	0.0194085
Talk	Word	First	0.9503537	0.0077375
Talk	Word	Middle	0.8772173	0.0158522
Silent	Letter	First	0.9673555	0.0077649
Silent	Letter	Middle	0.9303876	0.0175880
Silent	Word	First	0.9641518	0.0071661
Silent	Word	Middle	0.9120436	0.0140493

E1B_graph_acc

Figure 1

ggarrange(E1B_graph_iksi,E1B_graph_acc)

First letter IKSIs

E1B_FL_iksis <- talk_type_E1B_data

E1B_FL_iksis <- E1B_FL_iksis %>%
  filter(accuracy == 1,
         iksis < 5000,
         LetterType == "First") %>%
  mutate(condition = paste(voice,linguistic_unit,sep="_")) %>%
  mutate(subject = as.factor(subject),
         condition = as.factor(condition)) %>%
  group_by(subject,condition) %>%
  summarize(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
            prop_removed = modified_recursive_moving(iksis)$prop_removed)

E1B_FL_aov_out <- aov(mean_iksi ~ condition + 
                    Error(subject/(condition)), E1B_FL_iksis)

knitr::kable(xtable(summary(E1B_FL_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	959182.5	68513.036	NA	NA
condition	4	457062.7	114265.677	14.49192	0
Residuals1	56	441547.9	7884.784	NA	NA

E1B_FL_apa_print <- apa_print(E1B_FL_aov_out)
E1B_FL_means <- model.tables(E1B_FL_aov_out,"means")

E1B_FL_iksi_table <- E1B_FL_iksis %>%
              group_by(condition) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))%>%
              mutate(voice = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,1]),
                     linguistic_unit = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,2]))

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

E1B_FL_graph_iksi <- ggplot(E1B_FL_iksi_table, aes(x=linguistic_unit,
                                             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(100,500))+
  scale_y_continuous(minor_breaks=seq(200,500,25),
                     breaks=seq(100,500,100))+
    #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean IKSI (ms)")+
  theme(axis.title.x = element_blank())+
  facet_wrap(~voice, scales = "free_x", strip.position="bottom")

knitr::kable(E1B_FL_iksi_table)

condition	mIKSI	sem	voice	linguistic_unit
Normal_Normal	281.2607	32.40902	Normal	Normal
Silent_Letter	396.7084	30.10502	Silent	Letter
Silent_Word	277.8326	38.68840	Silent	Word
Talk_Letter	467.5775	38.24437	Talk	Letter
Talk_Word	279.4567	41.88154	Talk	Word

E1B_FL_graph_iksi

contrasts

# Normal vs. silentword and talk word
E1B_FL_NvsW <- apa_print(t_contrast_rm(df=E1B_FL_iksis,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "condition",
                             A_levels = c("Normal_Normal"),
                             B_levels = c("Silent_Word","Talk_Word"),
                             contrast_weights = c(-1,.5,.5) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -0.14099, df = 14, p-value = 0.8899
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -42.41293  37.18085
## sample estimates:
## mean of x 
## -2.616045

E1B_FL_NvsW

## $estimate
## [1] "$M = -2.62$, 95\\% CI $[-42.41$, $37.18]$"
## 
## $statistic
## [1] "$t(14) = -0.14$, $p = .890$"
## 
## $full_result
## [1] "$M = -2.62$, 95\\% CI $[-42.41$, $37.18]$, $t(14) = -0.14$, $p = .890$"
## 
## $table
## NULL

# Normal AND words VS. Letters

E1B_FL_NWvsL <- apa_print(t_contrast_rm(df=E1B_FL_iksis,
                                        subject = "subject",
                                        dv = "mean_iksi",
                                        condition = "condition",
                                        A_levels = c("Normal_Normal","Silent_Word","Talk_Word"),
                                        B_levels = c("Silent_Letter","Talk_Letter"),
                                        contrast_weights = c(-1/3,-1/3,-1/3,.5,.5)))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 4.7289, df = 14, p-value = 0.0003231
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   83.40268 221.84998
## sample estimates:
## mean of x 
##  152.6263

E1B_FL_NWvsL

## $estimate
## [1] "$M = 152.63$, 95\\% CI $[83.40$, $221.85]$"
## 
## $statistic
## [1] "$t(14) = 4.73$, $p < .001$"
## 
## $full_result
## [1] "$M = 152.63$, 95\\% CI $[83.40$, $221.85]$, $t(14) = 4.73$, $p < .001$"
## 
## $table
## NULL

# silent letter vs. talk letter
E1B_FL_SLvsTL <- apa_print(t_contrast_rm(df=E1B_FL_iksis,
                                        subject = "subject",
                                        dv = "mean_iksi",
                                        condition = "condition",
                                        A_levels = c("Silent_Letter"),
                                        B_levels = c("Talk_Letter"),
                                        contrast_weights = c(-1,1) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 2.2403, df = 14, p-value = 0.04181
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##    3.020064 138.718008
## sample estimates:
## mean of x 
##  70.86904

E1B_FL_SLvsTL

## $estimate
## [1] "$M = 70.87$, 95\\% CI $[3.02$, $138.72]$"
## 
## $statistic
## [1] "$t(14) = 2.24$, $p = .042$"
## 
## $full_result
## [1] "$M = 70.87$, 95\\% CI $[3.02$, $138.72]$, $t(14) = 2.24$, $p = .042$"
## 
## $table
## NULL

First letter Accuracy

E1B_FL_acc <- talk_type_E1B_data

E1B_FL_acc <- E1B_FL_acc %>%
  filter(
         LetterType == "First") %>%
  mutate(condition = paste(voice,linguistic_unit,sep="_")) %>%
  mutate(subject = as.factor(subject),
         condition = as.factor(condition)) %>%
  group_by(subject,condition) %>%
  summarize(mean_acc = mean(letter_accuracy))

E1B_FL_acc_aov_out <- aov(mean_acc ~ condition + 
                    Error(subject/(condition)), E1B_FL_acc)

knitr::kable(xtable(summary(E1B_FL_acc_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.0310106	0.0022150	NA	NA
condition	4	0.0093479	0.0023370	5.609321	0.00072
Residuals1	56	0.0233309	0.0004166	NA	NA

E1B_FL_acc_apa_print <- apa_print(E1B_FL_acc_aov_out)
E1B_FL_acc_means <- model.tables(E1B_FL_acc_aov_out,"means")

E1B_FL_acc_table <- E1B_FL_acc %>%
              group_by(condition) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc)))%>%
              mutate(voice = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,1]),
                     linguistic_unit = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,2]))

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

E1B_FL_acc_graph <- ggplot(E1B_FL_acc_table, aes(x=linguistic_unit,
                                             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(.8,1))+
  scale_y_continuous(minor_breaks=seq(.8,1,.05),
                     breaks=seq(.8,1,.05))+
    #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean Accuracy")+
  theme(axis.title.x = element_blank())+
  facet_wrap(~voice, scales = "free_x",
             strip.position="bottom")

knitr::kable(E1B_FL_acc_table)

condition	mAcc	sem	voice	linguistic_unit
Normal_Normal	0.9558603	0.0073141	Normal	Normal
Silent_Letter	0.9673809	0.0077634	Silent	Letter
Silent_Word	0.9641634	0.0071681	Silent	Word
Talk_Letter	0.9829568	0.0058096	Talk	Letter
Talk_Word	0.9503537	0.0077375	Talk	Word

E1B_FL_acc_graph

contrasts

# Normal vs. silentword and talk word
E1B_FL_acc_NvsW <- apa_print(t_contrast_rm(df=E1B_FL_acc,
                             subject = "subject",
                             dv = "mean_acc",
                             condition = "condition",
                             A_levels = c("Normal_Normal"),
                             B_levels = c("Silent_Word","Talk_Word"),
                             contrast_weights = c(-1,.5,.5) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 0.19301, df = 14, p-value = 0.8497
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -0.01413963  0.01693617
## sample estimates:
##  mean of x 
## 0.00139827

E1B_FL_acc_NvsW

## $estimate
## [1] "$M = 0.00$, 95\\% CI $[-0.01$, $0.02]$"
## 
## $statistic
## [1] "$t(14) = 0.19$, $p = .850$"
## 
## $full_result
## [1] "$M = 0.00$, 95\\% CI $[-0.01$, $0.02]$, $t(14) = 0.19$, $p = .850$"
## 
## $table
## NULL

# Normal AND words VS. Letters

E1B_FL_acc_NWvsL <- apa_print(t_contrast_rm(df=E1B_FL_acc,
                                        subject = "subject",
                                        dv = "mean_acc",
                                        condition = "condition",
                                        A_levels = c("Normal_Normal","Silent_Word","Talk_Word"),
                                        B_levels = c("Silent_Letter","Talk_Letter"),
                                        contrast_weights = c(-1/3,-1/3,-1/3,.5,.5)))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 3.5635, df = 14, p-value = 0.003117
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  0.007315949 0.029436839
## sample estimates:
##  mean of x 
## 0.01837639

E1B_FL_acc_NWvsL

## $estimate
## [1] "$M = 0.02$, 95\\% CI $[0.01$, $0.03]$"
## 
## $statistic
## [1] "$t(14) = 3.56$, $p = .003$"
## 
## $full_result
## [1] "$M = 0.02$, 95\\% CI $[0.01$, $0.03]$, $t(14) = 3.56$, $p = .003$"
## 
## $table
## NULL

# silent letter vs. talk letter
E1B_FL_acc_SLvsTL <- apa_print(t_contrast_rm(df=E1B_FL_acc,
                                        subject = "subject",
                                        dv = "mean_acc",
                                        condition = "condition",
                                        A_levels = c("Silent_Letter"),
                                        B_levels = c("Talk_Letter"),
                                        contrast_weights = c(-1,1) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 2.9643, df = 14, p-value = 0.01025
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  0.004306324 0.026845565
## sample estimates:
##  mean of x 
## 0.01557594

E1B_FL_acc_SLvsTL

## $estimate
## [1] "$M = 0.02$, 95\\% CI $[0.00$, $0.03]$"
## 
## $statistic
## [1] "$t(14) = 2.96$, $p = .010$"
## 
## $full_result
## [1] "$M = 0.02$, 95\\% CI $[0.00$, $0.03]$, $t(14) = 2.96$, $p = .010$"
## 
## $table
## NULL

Middle letter IKSIs

E1B_ML_iksis <- talk_type_E1B_data

E1B_ML_iksis <- E1B_ML_iksis %>%
  filter(accuracy == 1,
         iksis < 5000,
         LetterType == "Middle") %>%
  mutate(condition = paste(voice,linguistic_unit,sep="_")) %>%
  mutate(subject = as.factor(subject),
         condition = as.factor(condition)) %>%
  group_by(subject,condition) %>%
  summarize(mean_iksi = mean(modified_recursive_moving(iksis)$restricted),
            prop_removed = modified_recursive_moving(iksis)$prop_removed)

E1B_ML_aov_out <- aov(mean_iksi ~ condition + 
                    Error(subject/(condition)), E1B_ML_iksis)

knitr::kable(xtable(summary(E1B_ML_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	92054.71	6575.337	NA	NA
condition	4	172840.12	43210.031	19.12982	0
Residuals1	56	126491.63	2258.779	NA	NA

E1B_ML_apa_print <- apa_print(E1B_ML_aov_out)
E1B_ML_means <- model.tables(E1B_ML_aov_out,"means")

E1B_ML_iksi_table <- E1B_ML_iksis %>%
              group_by(condition) %>%
              summarize(mIKSI = mean(mean_iksi),
                        sem = sd(mean_iksi)/sqrt(length(mean_iksi)))%>%
              mutate(voice = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,1]),
                     linguistic_unit = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,2]))

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

E1B_ML_graph_iksi <- ggplot(E1B_ML_iksi_table, aes(x=linguistic_unit,
                                             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(100,500))+
  scale_y_continuous(minor_breaks=seq(100,500,25),
                     breaks=seq(100,500,100))+
   #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean IKSI (ms)")+
  theme(axis.title.x = element_blank())+
  facet_wrap(~voice, scales = "free_x",
             strip.position="bottom")

knitr::kable(E1B_ML_iksi_table)

condition	mIKSI	sem	voice	linguistic_unit
Normal_Normal	159.4457	9.458835	Normal	Normal
Silent_Letter	246.0493	18.117573	Silent	Letter
Silent_Word	156.7808	9.640095	Silent	Word
Talk_Letter	261.0719	20.618758	Talk	Letter
Talk_Word	152.2636	10.242864	Talk	Word

E1B_ML_graph_iksi

contrasts

# Normal vs. silentword and talk word
E1B_ML_NvsW <- apa_print(t_contrast_rm(df=E1B_ML_iksis,
                             subject = "subject",
                             dv = "mean_iksi",
                             condition = "condition",
                             A_levels = c("Normal_Normal"),
                             B_levels = c("Silent_Word","Talk_Word"),
                             contrast_weights = c(-1,.5,.5) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = -3.6742, df = 14, p-value = 0.002502
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -7.797614 -2.049506
## sample estimates:
## mean of x 
##  -4.92356

E1B_ML_NvsW

## $estimate
## [1] "$M = -4.92$, 95\\% CI $[-7.80$, $-2.05]$"
## 
## $statistic
## [1] "$t(14) = -3.67$, $p = .003$"
## 
## $full_result
## [1] "$M = -4.92$, 95\\% CI $[-7.80$, $-2.05]$, $t(14) = -3.67$, $p = .003$"
## 
## $table
## NULL

# Normal AND words VS. Letters

E1B_ML_NWvsL <- apa_print(t_contrast_rm(df=E1B_ML_iksis,
                                        subject = "subject",
                                        dv = "mean_iksi",
                                        condition = "condition",
                                        A_levels = c("Normal_Normal","Silent_Word","Talk_Word"),
                                        B_levels = c("Silent_Letter","Talk_Letter"),
                                        contrast_weights = c(-1/3,-1/3,-1/3,.5,.5)))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 4.8561, df = 14, p-value = 0.0002543
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##   54.38012 140.41441
## sample estimates:
## mean of x 
##  97.39726

E1B_ML_NWvsL

## $estimate
## [1] "$M = 97.40$, 95\\% CI $[54.38$, $140.41]$"
## 
## $statistic
## [1] "$t(14) = 4.86$, $p < .001$"
## 
## $full_result
## [1] "$M = 97.40$, 95\\% CI $[54.38$, $140.41]$, $t(14) = 4.86$, $p < .001$"
## 
## $table
## NULL

# silent letter vs. talk letter
E1B_ML_SLvsTL <- apa_print(t_contrast_rm(df=E1B_ML_iksis,
                                        subject = "subject",
                                        dv = "mean_iksi",
                                        condition = "condition",
                                        A_levels = c("Silent_Letter"),
                                        B_levels = c("Talk_Letter"),
                                        contrast_weights = c(-1,1) ))

## 
##  One Sample t-test
## 
## data:  contrast_vector
## t = 0.99829, df = 14, p-value = 0.3351
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -17.25303  47.29828
## sample estimates:
## mean of x 
##  15.02262

E1B_ML_SLvsTL

## $estimate
## [1] "$M = 15.02$, 95\\% CI $[-17.25$, $47.30]$"
## 
## $statistic
## [1] "$t(14) = 1.00$, $p = .335$"
## 
## $full_result
## [1] "$M = 15.02$, 95\\% CI $[-17.25$, $47.30]$, $t(14) = 1.00$, $p = .335$"
## 
## $table
## NULL

Middle letter Accuracy

E1B_ML_acc <- talk_type_E1B_data

E1B_ML_acc <- E1B_ML_acc %>%
  filter(
         LetterType == "Middle") %>%
  mutate(condition = paste(voice,linguistic_unit,sep="_")) %>%
  mutate(subject = as.factor(subject),
         condition = as.factor(condition)) %>%
  group_by(subject,condition) %>%
  summarize(mean_acc = mean(letter_accuracy))

E1B_ML_acc_aov_out <- aov(mean_acc ~ condition + 
                    Error(subject/(condition)), E1B_ML_acc)

knitr::kable(xtable(summary(E1B_ML_acc_aov_out)))

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
Residuals	14	0.2132002	0.0152286	NA	NA
condition	4	0.0354622	0.0088656	6.192274	0.0003373
Residuals1	56	0.0801759	0.0014317	NA	NA

E1B_ML_acc_apa_print <- apa_print(E1B_ML_acc_aov_out)
E1B_ML_acc_means <- model.tables(E1B_ML_acc_aov_out,"means")

E1B_ML_acc_table <- E1B_ML_acc %>%
              group_by(condition) %>%
              summarize(mAcc = mean(mean_acc),
                        sem = sd(mean_acc)/sqrt(length(mean_acc))) %>%
              mutate(voice = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,1]),
                     linguistic_unit = as.character(t(data.frame(strsplit(as.character(condition),split="_")))[,2]))

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

E1B_ML_acc_graph <- ggplot(E1B_ML_acc_table, aes(x=linguistic_unit,
                                             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(.8,1))+
  scale_y_continuous(minor_breaks=seq(.8,1,.05),
                     breaks=seq(.8,1,.05))+
  #scale_x_discrete(labels = c('Normal',
  #                            'Letter',
  #                            'Word',
  #                            'Letter',
  #                            'Word'))+
  ylab("Mean Accuracy")+
  theme(axis.title.x = element_blank())+
  facet_wrap(~voice, scales = "free_x",
             strip.position="bottom")

knitr::kable(E1B_ML_acc_table)

condition	mAcc	sem	voice	linguistic_unit
Normal_Normal	0.8915914	0.0161666	Normal	Normal
Silent_Letter	0.9298702	0.0176047	Silent	Letter
Silent_Word	0.9120436	0.0140493	Silent	Word
Talk_Letter	0.9336213	0.0194183	Talk	Letter
Talk_Word	0.8772173	0.0158522	Talk	Word

E1B_ML_acc_graph

ggarrange(E1B_FL_graph_iksi,E1B_ML_graph_iksi, 
          E1B_FL_acc_graph, E1B_ML_acc_graph, nrow=2, ncol=2,
          labels = list(A="First Letter ",
                        B="Middle Letters"),
          hjust=-.8)

speed accuracy tradeoff

library(ggrepel)
E1B_iksi_both <- rbind(E1B_FL_iksi_table,
                       E1B_ML_iksi_table)
E1B_iksi_both <- cbind(E1B_iksi_both, 
                       Letter_Position = rep(c("First","Middle"), each=5))

E1B_acc_both <- rbind(E1B_FL_acc_table,
                       E1B_ML_acc_table)
E1B_acc_both <- cbind(E1B_acc_both, 
                       Letter_Position = rep(c("First","Middle"), each=5))

E1B_SA <- cbind(E1B_iksi_both, accuracy = E1B_acc_both$mAcc)

E1B_SA_graph <- ggplot(E1B_SA, aes(x=mIKSI,  y=accuracy,
                   shape=Letter_Position,
                   color=Letter_Position,
                   label=condition))+
  geom_point()+
  geom_text_repel(size=1.7, color="black")+
  coord_cartesian(xlim=c(100,500), ylim=c(.8,1))+
  scale_x_continuous(breaks=seq(100,500,100))+
  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")
  

E1B_SA_graph

library(gridExtra)

grid.arrange(
  ggarrange(E1B_FL_graph_iksi,E1B_ML_graph_iksi, 
            E1B_FL_acc_graph, E1B_ML_acc_graph, nrow=2, ncol=2,
            labels = list(A="First Letter ",
                          B="Middle Letters"),
            font.label = list(size=10),
            hjust=-.8),
  E1B_SA_graph,
  ncol = 2, 
  widths=c(4.5,2.375))                       

#levels(E1B_FL_iksis$condition)
# normal, silent_letter, silent_word, talk_letter, talk_word
# set up linear contrasts
c1 <- c(1, 0, -.5, 0, -.5) # normal vs. silent_word & talk_word
c2 <- c(-1/3, .5, -1/3, .5, -1/3) # normal and word vs letter
c3 <- c(0,-1,0,1,0) # talk letter vs silent letter

# create a contrast matrix
mat <- cbind(c1,c2,c3)

# assign the contrasts to the group
contrasts(E1B_FL_iksis$condition) <- mat

# run the ANOVA
aov_out <- aov(mean_iksi ~ condition +Error(subject/condition), E1B_FL_iksis)

E1B_FL_means <- model.tables(aov_out,"means")

# print the contrasts, add names for the contrasts
summary(aov_out, split=list("condition"=list("normal vs word"=1, 
                                                "normalword vs letter" = 2, 
                                                "talk letter vs silent letter"=3)),
            expand.split=TRUE) 

## t test contrasts

# normal vs. words
a<-E1B_FL_iksis[E1B_FL_iksis$condition=="Normal_Normal",]$mean_iksi
b1<-E1B_FL_iksis[E1B_FL_iksis$condition=="Silent_Word",]$mean_iksi
b2<-E1B_FL_iksis[E1B_FL_iksis$condition=="Talk_Word",]$mean_iksi

t.test((-1*a)+((.5*b1) + (.5*b2)))

t_contrast_rm(df=E1B_FL_iksis,
              subject = "subject",
              dv = "mean_iksi",
              condition = "condition",
              A_levels = c("Normal_Normal"),
              B_levels = c("Silent_Word","Talk_Word"),
              contrast_weights = c(-1,.5,.5) )

# normal and words vs. letters
a1<-E1B_FL_iksis[E1B_FL_iksis$condition=="Normal_Normal",]$mean_iksi
a2<-E1B_FL_iksis[E1B_FL_iksis$condition=="Silent_Word",]$mean_iksi
a3<-E1B_FL_iksis[E1B_FL_iksis$condition=="Talk_Word",]$mean_iksi
b1<-E1B_FL_iksis[E1B_FL_iksis$condition=="Silent_Letter",]$mean_iksi
b2<-E1B_FL_iksis[E1B_FL_iksis$condition=="Talk_Letter",]$mean_iksi

cs <- c(-1/3,-1/3,-1/3,.5,.5)
c_vec <- (cs[1]*a1 + cs[2]*a2 + cs[3]*a3) + (cs[4]*b1 + cs[5]*b2)
t.test(c_vec)

t_contrast_rm(df=E1B_FL_iksis,
              subject = "subject",
              dv = "mean_iksi",
              condition = "condition",
              A_levels = c("Normal_Normal","Silent_Word","Talk_Word"),
              B_levels = c("Silent_Letter","Talk_Letter"),
              contrast_weights = c(-1/3,-1/3,-1/3,.5,.5) )

# silent letter vs. talk letter
a<-E1B_FL_iksis[E1B_FL_iksis$condition=="Silent_Letter",]$mean_iksi
b<-E1B_FL_iksis[E1B_FL_iksis$condition=="Talk_Letter",]$mean_iksi

t.test((-1*a)+(1*b))

t_contrast_rm(df=E1B_FL_iksis,
              subject = "subject",
              dv = "mean_iksi",
              condition = "condition",
              A_levels = c("Silent_Letter"),
              B_levels = c("Talk_Letter"),
              contrast_weights = c(-1,1) )

Save all

save.image(file="E1B_workspace.RData")

Factorial write-up

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")