If they share items (3-5), IRT score them for linking/equating for cut-off comparability.

Thanks! I'll check up that! Do you have a more detailed description or know an article who have used this procedure?

Best

jk

The answer may depend on your use of the scales.

Is this for an empirical study, and if so, is different groups measured with the 10 and 25-itemversions, or is the same persons measured at different points in time. Or is it a clinical problem with some patients being measured with each of the two versions, or alternatively the same persons measured at different points in time, and you want to know the change for these persons?

For clinical purposes, you may need the observed scores on the tests, and the recommendation of Matthew Kerry would apply. I am sure, there are a lot of empirical studies with examples of this. A general explanation can be found in de Ayala, R.J. (2009). The theory and practice of item response theory. New York: Guilford, chapter 11, and in Embretson and Reise (2000) Item response theory for psychologists. Mawah: Erlbaum. p.252ff. de Ayala describes three different approaches, Embretson and Reise one of them.

For research you may not need the observed scores (sum scores) at all, since by using SEM-methods you can include the measurement model and the structural model simultaneously. Measurement invariance would probably be an issue and should be investigated, however.

Thanks, Jan! It's about comparing two groups who score on the same scales, but when it came to psychiatric distress, SCL-10 was used in one group and SCL-25 in the other group. Both instruments are short versions of Hopkins Synptoms Check list, uses they same scale 1 - 4, have the same subfactors (anxiety & depression), but use different items for each factor.

jk

Hi John-Kåre

Then I would use SEM-type methods. This will give you more correct answers, and - importantly - it will probably show higher effects (if there in fact is effect) than using the traditional methods based on observed scores. This is because the traditional sum score measures are attenuated by the measurement error variance/unique item variance (so it is not a trick, the SEM-analyses are the correct ones).

Basically, I think you should use a measurement model based on the 10 SCL-items which are the same in the two versions of the test. Maybe you could treat the other 15 as missing data in the group evaluated with SCL-10. I am not sure if you will get any benefits from this. It might be.

You should check these 10 items for measurement invariance across the two groups, which also can be done with SEM-software.

As I don't know if you are familiar with SEM-methodology, I just want to recommend the program Mplus, and the wealth of learning materials you will find on the website for the program, www.statmodel.com. The program is not as costly as for instance SPSS. If price is a problem, however, there exists free programs, like lavaan and OpenMx. Both of these are packages for use with the statistical program R. This makes them more complicated to use, and they do not have the excellent at the same time simplicity and facility you will find in Mplus.

Best wishes

Jan

PS. As you are Norwegian, you might benefit of some of the information I have put on the web site www.forskningsmetode.dk -> Metoderessourcer, which however is still under construction.

Takk Jan! Vil sjekke ut dine forslag. Hilsen fra Norge!!

It's also worth noting that a number of studies with the SCL-90 item pool, which includes these items, have shown that the items and scales do not function equally across different populations (factor structure changes, items function differently). Depending on what sort of groups you are comparing, it might be important to look into this. I've found these papers useful on the topic:

Olsen, L. R., Mortensen, E. L., & Bech, P. (2004). The SCL-90 and SCL-90R versions validated by item response models in a Danish community sample. Acta Psychiatrica Scandinavica, 110(3), 225-229, doi:10.1111/j.1600-0447.2004.00399.x.

Paap, M. C., Meijer, R. R., Van Bebber, J., Pedersen, G., Karterud, S., Hellem, F. M., et al. (2011). A study of the dimensionality and measurement precision of the SCL-90-R using item response theory. International Journal of Methods in Psychiatric Research, 20(3), e39-55, doi:10.1002/mpr.347.

Paap, M. C. S., Meijer, R. R., Cohen-Kettenis, P. T., Richter-Appelt, H., de Cuypere, G., Kreukels, B. P. C., et al. (2012). Why the factorial structure of the SCL-90-R is unstable: Comparing patient groups with different levels of psychological distress using Mokken Scale Analysis. Psychiatry Research, 200(2), 819-826, doi:10.1016/j.psychres.2012.03.012.