DataWrangleFlankerAV.Rmd

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, warning=FALSE, message = FALSE)
```

## Flanker task


In a flanker task, participants identify a central stimulus (as quickly and accurately) as possible, while ignoring distracting stimuli presented on the left or right of the central stimulus (the flankers). 

The data for this assignment come from a flanker task where participants responded to many flanker stimuli over several trials. 

## Load the data and libraries you will use


```{r}

library(data.table)
library(dplyr)
library(ggplot2)
library(bit64)


# get the file names
file_names <- list.files(path="FlankerData")

# create headers for each column
the_headers <- c("stimulus","congruency","proportion",
                 "block","condition","dualtask","unknown",
                 "stimulus_onset","response_time","response","subject")
# Load data
# create empty dataframe
all_data<-data.frame()

# loop to add each file to the dataframe
for(i in file_names){
  one_subject <- fread(paste("FlankerData/",i, sep=""))
  names(one_subject) <- the_headers
  one_subject$subject <- rep(i,dim(one_subject)[1])
  one_subject <- cbind(one_subject, trial= 1:dim(one_subject)[1])
  all_data <- rbind(all_data,one_subject)
}

```

## Pre-processing

Create an accuracy column that codes whether the response was correct or incorrect on each trial 

```{r}
center_let<-unlist(lapply(strsplit(all_data$stimulus,""),
                         FUN=function(x)unlist(x)[2]))

all_data<-cbind(all_data,center_let)

all_data<-all_data %>%
    mutate(response=tolower(response),
    center_let=tolower(center_let),
    accuracy=response==center_let)
  

```

### Get Reaction time on each trial

 Add a column that calculates the reaction time on each trial.


```{r}

# Somehow the stimulus and the response time are the same number?

all_data <- all_data %>%
              mutate(RT = as.integer(response_time - stimulus_onset))

```


## Checks

Check how many trials each subject completed in the congruent and incongruent conditions, the mean accuracy for each subject in each congruency condition, and the mean RT for each subject in each congruency condition.

```{r}

Sub_Trials <- all_data %>%
                  mutate(subject = as.factor(subject),
                         congruency = as.factor(congruency)) %>%
                  group_by(subject,congruency) %>%
                  summarise(num_trials = length(RT),
                            mean_RT = mean(RT),
                            mean_accuracy = mean(accuracy))
knitr::kable(Sub_Trials)
```



### Exclusion

Exclude all RTs that are longer than 2000 ms

```{r}

# Reaction Time =0 somehow the stimulus and response are the same number
Slow_RT <- all_data %>%
              filter(RT < 2000)
```

## Analysis

### Reaction Time analysis

1. Get the individual subject mean reaction times for **correct** congruent and incongruent trials.

```{r}
Subject_Trials_meanRTs <- Slow_RT %>%
                    filter(accuracy == TRUE) %>%
                    group_by(subject,congruency) %>%
                    summarise(mean_RT = mean(RT))
```

2. Get the overall mean RTs and SEMs (standard errors of the mean) for the congruent and incongruent condition. Make a table and graph.

```{r}

 congruency_means <- Subject_Trials_meanRTs %>%
                      group_by(congruency) %>%
                      summarise(meanRT = mean(mean_RT),
                                SEM = sd(mean_RT)/sqrt(length(mean_RT)))
knitr::kable(congruency_means)

ggplot(congruency_means, aes(x=congruency,y=meanRT,
                             fill=congruency))+
  geom_bar(stat="identity")+
  theme_classic(base_size=12)+
  ylab("Mean Reaction Time ")+
  geom_errorbar(aes(ymin=meanRT-SEM,
                    ymax=meanRT+SEM),
                position=position_dodge(width=0.9),
                width=.2,
                color="black")+
  coord_cartesian(ylim=c(100,1000))

  
```

3. Compute the flanker effect for each subject, taking the difference between their mean incongruent and congruent RT. Then plot the mean flanker effect, along with the SEM of the mean flanker effect

**tip:** Not all problems have an easy solution in dplyr, this is one them. You may have an easier time using logical indexing of the dataframe to solve this part.

```{r}
library(ggplot2)
flanker_means <- subject_meanRTs[subject_meanRTs$congruency=="I",]$mean_RT - 
  subject_meanRTs[subject_meanRTs$congruency=="C",]$mean_RT
flanker_df <- data.frame(dv = "flanker effect",
                         flanker_mean = mean(flanker_means),
                         SEM = sd(flanker_means)/sqrt(length(flanker_means)))
ggplot(flanker_df, aes(x=dv, y=flanker_mean))+
  geom_bar(stat="identity")+
  theme_classic(base_size=12)+
  ylab("Mean Flanker Effect")+
  geom_errorbar(aes(ymin=flanker_mean-SEM,
                    ymax=flanker_mean+SEM),
                position=position_dodge(width=0.9),
                width=.2,
                color="black")
  
```


### Exploratory analysis

Multiple questions may often be asked of data, especially questions that may not have been of original interest to the researchers. 

In flanker experiments, like this one, it is well known that the flanker effect is modulated by the nature of the previous trial. Specifically, the flanker effect on trial n (the current trial), is larger when the previous trial (trial n-1) involved a congruent item, compared to an incongruent item. 

Transform the data to conduct a sequence analysis. The dataframe should already include a factor (column) for the congruency level of trial n. Make another column that codes for the congruency level of trial n-1 (the previous trial). This creates a 2x2 design with trial n congruency x trial n-1 congruency. 

First get teh subject means for each condition, then create a table and plot for teh overall means and SEMs in each condition. This should include:

1. trial n congruent : trial n-1 congruent
2. trial n incongruent : trial n-1 congruent
3. trial n congruent : trial n-1 incongruent
4. trial n incongruent : trial n-1 incongruent

**tip:** be careful, note that the first trial in each experiment can not be included, because it had no preceding trial

```{r}
library(dplyr)
previous_congruency <- c(0,all_data$congruency[1:(length(all_data$congruency)-1)])
sequence_subjects <- all_data %>%
                      mutate(n1_congruency = previous_congruency) %>%
                      filter(trial > 1,
                             RT < 2000,
                             accuracy == TRUE) %>%
                      group_by(subject,congruency,n1_congruency) %>%
                      summarise(meanRT = mean(RT))
sequence_means <- sequence_subjects %>%
                    group_by(congruency,n1_congruency) %>%
                    summarise(mean_RT = mean(meanRT),
                              SEM = sd(meanRT)/sqrt(length(meanRT)))
                      
knitr::kable(sequence_means)
ggplot(sequence_means, aes(x=n1_congruency, y=mean_RT,
                           group=congruency,
                           fill=congruency))+
  geom_bar(stat="identity", position="dodge")+
  theme_classic(base_size=12)+
  ylab("Mean RT (ms")+
  geom_errorbar(aes(ymin=mean_RT-SEM,
                    ymax=mean_RT+SEM),
                position=position_dodge(width=0.9),
                width=.2,
                color="black")+
  coord_cartesian(ylim=c(600,900))

```

[link to Presentation](https://13vasquezan.github.io/RClassWeb/Presentation-Final#1)