You are here

High MZ correlations exceed the Chronbach's alpha of the Questionnaire

3 posts / 0 new
Last post
Moritz's picture
Offline
Joined: 02/10/2015 - 11:44
High MZ correlations exceed the Chronbach's alpha of the Questionnaire

Hi Everyone,

I have a more general question about MZ and DZ correlations and their relationship with the Chronbach's alpha of the questionnaire that is used.

I am running different twin models to estimates A,C and E contribution to variation in behaviours using a parent measured questionnaire in a sample of infants and toddlers.

When running the analyses I am consistently finding large MZ correlations (0.80-0.95) for the different traits, which exceed the chronbach's alpas of the scale.
To my knowledge the Chronbach's alpha provide an upper limit to the ICCs and hence the ICCs should not be higher than the alpha.
All traits are normally distributed and substantial differences between MZ and DZ pairs are being found, so the fact that parents just rate their twins the same due to the young age of sample cannot explain this completely.

Basically: If the MZ correlation is consistently higher than the Cronbach’s alpha for many of the psychometric measures, does this mean there is systematic bias in the data?

What do you think?
Thank you for your help
M

AdminRobK's picture
Offline
Joined: 01/24/2014 - 12:15
Same sample?

Were the Cronbach's alphas computed from the same twin sample in which you're calculating intraclass correlations?

Edit: also, do you know anything about the latent dimensionality of the questionnaires? Like, would you extract more than one factor in a factor analysis of each questionnaire's items?

Moritz's picture
Offline
Joined: 02/10/2015 - 11:44
Hi there Thanks for your

Hi there

Thanks for your reply.
Yes the alphas were calculated from the same sample as the ICCs.
Atm, my only explanation is that some of the scales only consists of 3 items, which will limit the value of the chronbach alpha and hence lead to the difference between ICCs and alpha.

I do not have the details on the factor analysis used to group the items, so cannot give you more details on the factor structure at this point.

Thanks
M