If a questionnaire includes positively and negatively-keyed items,
1)should all the negatively keyed items be reverse coded?
for example, if there are two items, 1. I like the game 2. The game is boring, the 2nd one can be reverse coded. But, Disagreement to “game is boring” is logically similar to “I like the game" .
2) Should i consider both the statements for reliability test? should i omit one item? Can anyone help me to clarify this doubt?
3) And the other doubt is, item 3. is " I could not do the game well". Should it be reverse coded? There is no positively worded statement for this item.
4) I could not find literature to clarify my doubts..Can anyone help me to clarify my doubts?
Thanks in advance !!!
1. Reverse all negatively keyed items.
2. Include all items in the reliability tests after recoding.
3. If you are unsure whether or not an item should be reverse coded, you can test it by including it in the reliability test both ways, each one at a time, to see which is best. An exploratory factor analysis will also confirm that all items belong in the measure.
You do not need to change the wording, just the responses (if the Likert scale is 1 to 5 then
(reverse coded response) = 6 - (original response)
There are two reasons for doing this: the Cronbach Alpha would be a more appropriate measure of consistency, and if you calculate factor scores then the inverse coded items would potentially cancel out the positively coded items and all the factor scores could be close to 3. In SPSS, when you calculate the Cronbach alpha, there is an option to ask for each item what would alpha be if that item were deleted. More directly, if you do a Principal Components Analysis, the inverse coded items will load negatively on the principal components, which would suggest that reverse coding is needed.
The reason that positively and negatively worded items are included in an assessment is to counteract positivity and negativity biases among respondents. Look for literature discussing those two constructs. Of course when you are measuring a variable, many times you are assuming unidimensionality with values extending from "less" to "more" of a trait or quality that you are measuring. However, in order to avoid the problem of people simply choosing high or low values and therefore biasing the results, you include some items in which the quality is measured the other way round (from "more" to "less"). During data analysis, you recode those items so that once more the entire scale is unidimensional.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire