Get your dataset ready

Upload, clean, restructure and save data

Get R and youR data Ready

Get ready

To work with R, you need packages, including useful functions. These will save time because they provide an easy way to run complex analyses. By installing these packages on your computer, you can make use of these functions. Indeed, they are already written for you, so you don’t have to code yourself.

Tip

At each start of a new R session (e.g., after shutting down your computer), you have to load the packages you want to use. Instead of using library(X) for each package, you can make a list so they can be loaded all at once.

Following packages are required for (1), (2) and (3):

# my list of packages (as an example)
x<-c('readxl','haven','psych','sjPlot','writexl','foreign')

# do this only once
install.packages(x) 

# do this at the start of each session
lapply(x, require, character.only = TRUE) 

# use this for only one package
library(PACKAGE_NAME)

Also important is that R knowns in which folder you want to work. By following code, you set your work directory in which R can find the required datasets and also can save the files you want to save.

setwd("/Users/joachimwaterschoot/Downloads/Analyses")

Welcome, data!

Datasets can be loaded in every format. Beware, these functions also include more options (e.g., a specific tab in an excel file). More information can be found on the page of the packages (search for the function in Google)

df <- readRDS('FILE.rds')
df <- read_xlsx('FILE.xlsx')
df <- read.spss("FILE.sav",
                use.value.labels = FALSE,
                to.data.frame=TRUE)

After data is loaded, you can check whether it occurred correctly. Do this by checking the column names and the dimension (number of rows X number of columns).

names(df) #column names
dim(df) #how many rows and columns
head(df,10) #check the first 10 rows

WoRk with the dataset

Put your variables in the right format

Importantly, you want to have variables in the right format. This is important before starting analyses.

# formatting into numeric variable
df$VARIABLE1 <- as.numeric(df$VARIABLE1)

# formatting into a categorical variable
df$VARIABLE2 <- as.factor(df$VARIABLE2)
# check the levels
levels(df$VARIABLE2)
# provide different labels to the levels if necessary
levels(df$VARIABLE2) <- c('LEVEL1','LEVEL2')

How to reverse a variable?

To reverse a numeric value or to change specific values to another value in a column, you can use the recode function…

library(dplyr)
df$item_r <- as.factor(recode(df$item, '1'='5', '2'='4','3'='3','4'='2','5'='1'))

… or you can subtract that column from its highest value + 1.

Here, a 1-5 scale is reversed by subtracting it from 6. By doing this, 5 becomes 1, 4 becomes 2, etc.

df$item_r <- 6-as.numeric(df$item)

How to match a variable to another dataset?

When you want to add information to a dataset (e.g., df), based on another dataset (e.g., df_other), you can use the match() function to first match values of the same variable in the same datasets (e.g., participation number).

df$VARIABLE1 <- df_other$VARIABLE2[match(df$ID,df_other$df)]

How to rename a column?

names(df)[names(df) == 'old.var.name'] <- 'new.var.name'

Create variables

In the keys.list, you make an overview of which items belong to which variable. Do this for only those containing at least 2 items or more.

Important

Of course, this can only be done with numeric variables.

keys.list <- list(
  VARIABLE1 = c("ITEM1","ITEM2"),
  VARIABLE2 = c("ITEM1","ITEM2")
)

This list will be used in the following code to calculate your variables and to add them in the described order to the dataset.

How to rename a column?

In this code, the name of the dataframe is df. Make sure you replace this label when you have another name

library(psych)

columns <- unlist(keys.list, use.names=FALSE) 
scaleitems <- df[,columns] 
scaleitems <- sapply(scaleitems, as.numeric) 
df <- df[, ! names(df) %in% columns, drop = F]

keys <- make.keys(scaleitems,keys.list)
scores <- scoreFast(keys, scaleitems, impute="none")

means <- as.data.frame(scores)
colnames(means) <- sub("-A.*", "", colnames(means))

df <- cbind(df, scaleitems,means)

The sjt.itemanalysis function of the sjPlot package provides a nice overview of the internal consistencies of your variables.

Note

• overview of items
• percentage of missing values
• standard deviation
• skewness: the higher, the skewer
• item difficulty: should range between .20 and .80. Ideal value is p+(1-p)/2 (mostly between .50 and .80)
• item discrimination: acceptable cut-off of .20. The closer to 1, the better.
• Cronbach’s Alpha if item was removed from scale
• mean (or average) inter-item correlation: acceptable between .20 and .40
• Cronbach’s Alpha: acceptable cut-off of .70

sjt.itemanalysis(df[,c(keys.list$VARIABLE1)])
sjt.itemanalysis(df[,c(keys.list$VARIABLE2)])

Save a dataset

saveRDS(df,'NAME.rds') # R format
write_csv(df,'NAME.csv') # csv format
write_xlsx(df,'NAME.xlsx') # excel format
write_sav(df,'NAME.sav') # SPSS format

Long and wide

Wide to long

First, we make a list of those variables we want to restructure.

longlist <- list(
  VAR1=c('VAR1_1','VAR1_2','VAR1_3','VAR1_4'),
  VAR2=c('VAR2_1','VAR2_2','VAR2_3','VAR2_4')
)

The list is used in the code below:

dflong <- reshape(data = df, # name of the dataset
                  idvar = "ID", # group variable including dependent variance
                  varying = longlist,
                  direction="long",
                  v.names = names(longlist),
                  sep="_")

Long to wide

Complete the following function:

dfwide <- reshape(dflong, # name of the dataset
                  idvar = "ID",  # name of the grouping variable including dependent variance
                  timevar = "time", # time variable
                  direction = "wide")

head(dfwide) # check whether the function did what you expected