Hi,
I need help on how to increase the Cronbach's Alpha of my survey questionnaire.
The survey questionnaire uses a 7-point likert scale and I had 12 respondents for its pilot testing. The survey questionnaire has a Cronbach's Alpha of 0.75 (good reliability). Now that the actual surveying is done with 432 survey respondents, the survey questionnaire's Cronbach's Alpha decreased to 0.63 (fair reliability).
I am puzzled why that is the case and I cannot find any answer or reference to help me with this.
Please help.
Thank you in advance.
Dear Matthew Salang
If you want to increase the reliability, you can check the individual Cronbach's Alpha values of all the variables (items) and remove the variables which are creating low values. Use the "Cronbach's Alpha if item deleted" option in SPSS.
Refer this video
https://www.youtube.com/watch?v=Kz8OdR6lV44
I hope it helps. :)
Also, take a step back and think about if you expect alpha to be high. Should all the items be indicating the same latent concept ?
AI-GENERATED TEXT WARNING for Onipe Adabenege Yahaya
's answerhttps://quillbot.com/ai-content-detector
100% of the text classified as generated by AI.
So Onipe Adabenege Yahaya
doesn't know the answer, and has no way of knowing if what he copied and pasted is correct or not. Ignore.
You can increase Cronbach's alpha by asking the exact same questions 2 or 3 times. The question is why do you want to increase Cronbach's alpha.
Matthew, the first thing you should do is to test the dimensionality of your data, because Cronbach's alpha is based on the assumption that the items are unidimensional. Just assuming that the items are unidimensional is not enough. Therefore, you could perform an exploratory factor analysis to get an idea about the number of factors. As you seem to have a theory about the relationships of your items, you could also perform a confirmatory factor analysis and investigate the modification indices. These will tell you whether some error covariances should be added to the model.
Kind regards, Karin
Using alpha-if-item-removed is not useful here. It's a stepwise method and the subset of items you get are not in any way guaranteed to have the lowest alpha for that number of items.
Have a look at Mokken scaling, which is nonparametric, based on a strong definition of a unidimensional scale and can disentangle several subscales within an item pool.
I will take your answers into consideration.
Thank you very much for your support!
- Badges
- Science topic
- Similar topics
- Filing