# Cronbachs alpha

Home Forums Methodspace discussion Cronbachs alpha

Viewing 5 posts - 1 through 5 (of 5 total)
• Author
Posts
• #1806
Kay Smith
Member

I have calculated Cronbachs alpha for a recent study and have come up with a value of 0.9023 (raw) (though the stats package I use also gives a standardised value as well).  However, there is a paper published with a similair study to mine that then goes on to caluclate an “alpha with deleted variable” (for both raw and standardised value.

The paper doesn’t explain why and I have done a web (and text) search and cannot find the answer.

Does anyone have any input?

#1810
Dhruba Lahiri
Participant

Assuming you are using SPSS and if I understood your question correctly when you do a relaibility test under the ‘statistics’ option you can select the option ” scale if item deleted”.This will give you a new column which gives the Alpha if item is deleted.

Taking your example. your overall score is 0.9023 and you would expect that the individual alpha of each item would be below this. If any of them are above ( lets say 0.9123) that implies that if that item is deleted then you could achive an even higher overall alpha. This variable in question could be eligible for removal.

#1809
Kay Smith
Member

I am using NCSS which is a very similair package.

Your explanation below is exactly what i was looking for – thankyou so much.  Are you able to explain to me what a standardised vs raw score is?

#1808
Dhruba Lahiri
Participant

I am not too clear as to the purpose of a standardised alpha. However, a good book that will give some understanding of Relaibility analysis is ” Discovering Statistics using SPSS’ by Andy Field. Unfortunatley I have the second edition and the explanation of reliability analysis is given in Chapter 15 ( pp 666-676). As I understand from the write-up the normal alpha (raw) is calculated using covariances while thr standised score is calculated using correlations.

#1807
Kay Smith
Member

thanks

Viewing 5 posts - 1 through 5 (of 5 total)
• You must be logged in to reply to this topic.