I am doing a replication study on scales of the form

Y = a0+ a1X1 + a2X2 + a3X3

The scales were developed three decades ago. In my dataset I found Cronbach's Alpha between .8 and .9 when analysing the X scores.

My questions are:

  • Does it make sense to also Cronbach the aiXi terms? I was advised to do so, but I have my doubts.
  • If yes, does it make sense to multiply all ai  with - say - a factor 10 -and then do a Cronbach?
  • Or is it best to ignore Cronbach alltogether and conduct a CFA on my own data. I did so and found RMSEAs around .02, which is quite convincing that the factors still emerge. This is (more or less) suported by the alphas.on the Xi variables. However, all alphas derail completely if Cronbach is conducted on the aiXi terms. I found structural equivalence in a multiple groups analysis, but the path coeficients I found differ from the ai in the original equations.
  • I did not find much recent literasture on Cronbach, but I am aware that in handbooks on SEM this test is hardly promoted - or even completely ignored  (e.g. Stevens, Byrne, Brown). So, some references would be great as well.

    Thanks!

    Similar questions and discussions