4_data_seq_analysis.R

#######################################################################
# R code accompanying Sau-Chin Chen et al. (2019)
# OSF projet: https://osf.io/e428p/
# ---------------------------------------------------------
# Download rawdata from lab OSF
# Written by Sau-Chin Chen
# E-mail: pmsp96@gmail.com
# Last update: October 17, 2019
#############################################################

library(tidyverse)
library(data.table)
library(magrittr)

# Import multiple-bytes string in English system
Sys.setlocale("LC_ALL","English")

##set working directory to current project folder
setwd(here::here())

## Sequential analysis, subset by the end sequence
require(BayesFactor)

BF_sequential <- function(DF, acc, scale=0.707){

  data <- DF %>%
    group_by(subject_nr) %>%
    subset(correct == 1)
  tmp_seq <- data %>% pull(subject_nr) %>% table() %>% unlist()
  tmp_ind <- data.frame(id = rep(names(tmp_seq) %>% as.numeric(), tmp_seq), end_seq = rep(names(tmp_seq) %>% order(), tmp_seq))
  data <- data %>% left_join(tmp_ind,by=c("subject_nr"="id")) %>%
    group_by(end_seq, subject_nr, Match) %>%
    summarise(V_RT = median(response_time, na.rm = TRUE), V_Accuracy = (n()/12) )

  bfProg <- matrix(0, ncol = 3, nrow = length(unique(data$end_seq)) )
  colnames(bfProg) <- c("ID","N", "BF")

  for(i in data$end_seq){

    ## Initial Point
    if(i < 3){
      bfProg[i, "ID"] <- as.numeric(names(tmp_seq)[i])
      bfProg[i, "N"] <- 1
      bfProg[i, "BF"] <- 1
    }

    if(i > 2){
      ## BF analysis for large and small objects
      ## Save the current sample size
      bfProg[i, "ID"] <- as.numeric(names(tmp_seq)[i])
      bfProg[i, "N"] <- i

      bfProg[i, "BF"] <- (ttestBF(
        x = data %>%
          filter(end_seq %in% (1:i), Match == "N") %>%
          pull(V_RT),
        y = data %>%
          filter(end_seq %in% (1:i), Match == "Y") %>%
          pull(V_RT),
        nullInterval = c(0,Inf),
        paired = TRUE,
        rscale = scale) %>%
          ## Return BF10 and BF01
          extractBF(onlybf = TRUE))[1] #%>%
      #round(2)
    }


  }

  data.frame(bfProg)
}

# Processing lab data
rawdata_SP_V <- dirname(getwd()) %>%
                dir(pattern = "all_rawdata_SP_V",
                    recursive = TRUE, full.names = TRUE) %>% read.csv
# Mark Acc < .70
rawdata_SP_V = (rawdata_SP_V %>%
  group_by(PSA_ID, subject_nr) %>%
  summarise(Acc_bound = (mean(correct) < .7))) %>%
  right_join(rawdata_SP_V, by = c("PSA_ID","subject_nr"))

## Setup of randomization seed
set.seed(100)

## Run sequential analysis and export the result to lab folder
for(LAB in unique(rawdata_SP_V$PSA_ID)){
  ## Get the raw data from the storage path
  route <- (dirname(paste0(getwd(),"/1_raw_data/")) %>%
          dir(pattern = paste0(LAB,"$"),
            recursive = TRUE, full.names = TRUE, include.dirs = TRUE))[1]  ## Make sure only one target directory

  BF_sequential(subset(rawdata_SP_V, (PSA_ID == LAB & Acc_bound != TRUE) )) %>% ## Revise the condition in terms of the preregistered plan
    write.csv(file = paste0(route,"/Seq_output.csv"),row.names = FALSE)
  print(paste(LAB,"analyzed"))
}

## Export the rawdata_SP_V
write_csv(rawdata_SP_V, file = "1_raw_data/all_rawdata_SP_V.csv")