0
$\begingroup$

I have two questions:

First, I am adapting a measurement instrument (scale) from English to Turkish. The original scale has six factors. The results of an exploratory factor analysis in SPSS, demonstrated that the adapted instrument consists of nine factors. I deleted one factor, because its items had similar loadings on different factors.

Can the adapted scale be used as an instrument measuring eight factors?

Secondly, an assessment of the internal consistency demonstrated a high Cronbach's $\alpha$ for the total scale, but a very low $\alpha$ for two factors i.e., .200 and .387.

Should these factors be removed from the instrument?

$\endgroup$
  • $\begingroup$ A "scale" which is multi-factor in its structure should rather be called a "battery". Anyway, it may not be expected to be homogeneous, i.e. high alpha. $\endgroup$ – ttnphns Apr 3 '14 at 21:51
  • $\begingroup$ Actually PCA is a exploratory procedure, with not a great reliability when ratio N/i is small i.e. smaller than 10 (N: total respondents, i: total items in PCA procedure). Do you consider applying Confirmatory Factor Analysis instead? How many respondents do you have? If sample is large you may consider splitting the sample in two parts, apply PCA in first half and CFA in second according to the proposed model in original scale. $\endgroup$ – Epaminondas Apr 4 '14 at 5:52
2
$\begingroup$

Before you conclude that the factors are poor, check if any items are correlating negatively with the others. Factors will adjust for this automatically, but alpha will not.

Then, rather than trying to throw away whole factors, I would look at each item in each factor; examine its correlation. Check for poor item quality (e.g. everyone or almost everyone giving the same answer).

$\endgroup$
0
$\begingroup$

Since you are adapting an existing measurement instrument (rating scale), you have some theoretical knowledge about how each item should be related a specific scale i.e. you know the measurement model for the English version. Therefore, it makes sense to perform a confirmatory factor analysis (CFA) on the Turkish data to test the fit of that particular measurement model.

Subsequently, the internal consistency of the scales can be calculated using the coefficients of the CFA.

There are several statistical packages that can be used for CFA, including Mplus,lavaan in R,AMOS, and LISREL.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.