I'm not referring to validating the instrument because it's already been validated, but at examining the scores of a particular sample that did the survey.

You can use an oblique rotation on your factor analysis to see how strongly the factors are related to each other. But I would not recommend an item-to-total analysis because the total includes the items with which you would be correlating it.

Diane Rozells, if I understand your question correctly, I think you could use Pearson's correlation to address the issue you raise. Given that you have a number of items on each factor, as well as on the total across factors, you could assume that each variable is at the equal-interval, rather than ordinal, level of measurement. Obtaining a Pearson's correlation coefficient is pretty straightforward in SPSS.

Thanks, Robert Trevethan and David. Robert, yes, I was thinking of a simple Pearson's correlation analysis because they can be treated as interval data, but my doubt (which was raised by David) is whether I can treat the total score as another variable in the typical correlation procedure (Analyze->correlate->bivariate -> Pearson) even though it encompasses the other variables. Or whether there was something extra I had to do in the analysis because of that.

You can use stata software after collecting data and you can see the association from the p value, weather there is association or not

Diane Rozells, I'm glad you followed up on my post. I confess that I interpreted your question as being primarily and simply a "mechanical" one (i.e., how can I get a correlation between each factor and the total score?) rather than something that required conceptual analysis.

So, now I'd come back to suggest that you consider carefully why you want to do what you are proposing. Certainly, there is an issue (related to the kind of issue that David Morgan raised) in that a total score is made up of responses on items from different factors, so any correlations between individual factors' scores and the total score are "contaminated" because the total score contains scores on items from each of the individual factors. In a sense, you'd be double dipping in the analyses to some extent.

If you think that what you want to do is conceptually justifiable (I'm not sure it is . . . ), I suggest you obtain separate total scores for each of the five factors on the helplessness scale, and also five composite scores, each comprised of the totals from a different set of four factors. Then correlate the total scores on each factor with the appropriate composite score of the other four factors (i.e., whichever composite score does not include responses from the single factor in question). And I'd use Pearson's for that.

I hope that makes sense.

But before you go to that trouble, I suggest you run an EFA on the data you have - if you have enough data to conduct EFA with. In my experience, the factor structure of data often does not conform with the factor structure of a published scale. It might pay to become really familiar with the nature of your data before you go to the effort of correlating subsections within those data.

Do feel free to get back!

Hi. Simply rely on the factor score correlation matrix received after factor extraction with oblique rotation. Assuming the total (or mean) of factor scores along rows as a new variable may not be a good idea. A reliability coefficient is another type of inter-item correlation and may be compared across different sub-scales.

Hello. If you already know that these 5 items work as a single factor, then you can summarize them in an index (variable transformation). You standardize this new index and use this new unit variable to correlate with the total score, Pearson in this case if they are continuous. I still think that the amplitude of each vector will matter in order to correlate. Check that part.

You might consider what's called the "item-rest" (or "corrected item to total") correlation. Subtract the value of the item (or in your case, the group of items) from the total score on all items, and correlate the resulting variable with the item (group of items) of interest. You can do an internet search on the quoted terms for more information.

As others have noted, this only makes sense if all of the items (all factors or groups of items) are expected to be highly correlated because they are all believed to measure the same thing. Hence, the corrected item to total correlation is most commonly of interest when examined separately for the items within each factor, rather than altogether for the factors across all items.