0

I am performing the reliability analysis of a psychometric scale which measures user engagement during a video game play. The scale has 28 questions, out of which the ones highlighted in yellow in the screenshot below will need to be reverse coded in the output (since they are opposite to what the scale purports to measure). enter image description here

library(readxl)
GES <- read_excel("CAGE Datadump Copy.xlsx")
GESpreEngagement <- GES[,11:38] #these 28 columns contains the response to 28 questions from survey
GESpreEngagementReverseCOded <- GESpreEngagement

# data in column 9 through 16 needs to be reverse coded, corresponding to the questions highlighted in the screenshot
GESpreEngagementReverseCOded[,c(9:16)] = 5-GESpreEngagement[,c(9:16)]

#calculate the cronbach's alpha value
psych::alpha(GESpreEngagementReverseCOded, check.keys = TRUE)

This codes is leading to the following output with a warning message(with or without check.keys=TRUE):

In psych::alpha(GESpreEngagementReverseCOded, check.keys = TRUE) :
Some items were negatively correlated with total scale and were automatically reversed. This is indicated by a negative sign for the variable name.

enter image description here

However, there is no such warning if I do not reverse code the output, i.e. if I just run psych::alpha(GESpreEngagement). The question is, it seems logical to reverse code the output, but R is telling me to do otherwise. What should I do in this case?

Vipin Verma
  • 5,330
  • 11
  • 50
  • 92
  • check if the items are positively correlated after reversal, if it doesn't work as expected, I would suspect a semantic issue or an understanding issue. the yellow questions seems to me that they should be reversed. use the reverse.code from psych package to reverse – Dimitrios Zacharatos Sep 07 '21 at 06:27

1 Answers1

1

I also ran into this issue recently and it was driving me crazy. In my case I later found out that the raw data was already reverse-coded before I imported it, which is why reversing them again gave me this warning. This is pretty late but hope it's been sorted out!

F_Average
  • 11
  • 2