Different value of Cronbach's alpha with psych and ltm packages and SPSS

[benagoodman]

Hi,

I'm trying to look at the internal consistency of a questionnaire consisting of 9 binary items.
I mainly use SPSS, but am using R to determine McDonald's Omega coefficient using the omega() function in the "psych" package.

The main output I get from the omega function is:
Alpha: 0.71
G.6: 0.79
Omega Hierarchical: 0.58
Omega H asymptotic: 0.7
Omega Total 0.83

The value of alpha is markedly different to the alpha that SPSS gives me: 0.559
0.559 is also the value that the coefficient.alpha() function in the "ltm" package, so I'm inclined to believe that one more than 0.71

When reading the SPSS file in R, I get the following error message (although I read somewhere else that it was irrelevant to the statistical calculations):

In read.spss(datafilename, to.data.frame = TRUE, use.value.labels = false) :
Unrecognized record type 7, subtype 18 encountered in system file

Can anyone tell me why the omega function gives me a different value for alpha?
And should I trust the other omega outputs?
And is the error message relevant?

Many thanks for any help,
Ben

 

[Jake]

Just guessing, but the alpha from the omega() function might be the so-called standardized alpha, with the others being unstandardized (standardized alpha >= unstandardized alpha). Try manually standardizing each of your items and then running the 3 alpha functions again to see if they all agree.

 

[benagoodman]

Thanks Jake,
I thought that too. However the standardised alpha is 0.581.
I'm not sure what the omega function's alpha is calculating.
Any other suggestions?
Cheers,
Ben

 

[benagoodman]

Figured it out:
Using the alpha() function in the psych package I got the following error message:

Warning message:
In alpha(eli) :
Some items were negatively correlated with total scale and were automatically reversed

So it was monkeying around with the scale without me telling it to.

Putting flip=false on the omega() parameters made it all better:
Omega
Call: omega(m = eli, flip = FALSE)
Alpha: 0.58
G.6: 0.72
Omega Hierarchical: 0.51
Omega H asymptotic: 0.65
Omega Total 0.78