I am doing a systematic review looking at factors that could predict the level of burden in stroke carers. I have narrowed down to the final studies I want to use and I have their data. Unfortunatley the only data I have are the Means and SDs of the factors and burden levels. I am trying to work out how I can look at if there is a correlation or not, but I can't work out how to do it. Everything I try (except for SPSS summary independent samples t-test) doesn't seem to work, I think because it takes each result from a study as one participant and I cant find anyway of me saying for this study there were x participants and this as the mean result. Can anyone help?
Hello Julia,
The ideal way is to have either: (a) reported inter-variable correlation values for each individual study; or (b) access to the raw data from the individual studies. It might be that you could contact authors of the respective studies and request either (a) or (b). Barring that, your task will be quite difficult.
A proxy method, which could yield somewhat misleading results, would be to use the study as the individual case/unit/replication, and record the respective means as if they were raw data points. Given a sufficient number of studies, you could report on whether average IV values did or did not relate systematically to average DV values across studies. You could optionally weight the studies by their sample sizes so as to give more impact to the larger-N studies.
However, the potential danger is that, the behavior of inter-variable relationships is often different for individual cases than for batches of cases (which is why multi-level designs have become so important in the past 20-30 years).
Good luck with your work.
Yes, it is possible to make a correlation, as well as test the results using any statistical program such as SPSS or SAS and others
From your description it seems likely that you want to conduct some kind of meta-analysis. This isn't something I'd do via SPSS (though it is possible using macros). My own preference would be to use the metafor package in R (but there are other packages in R and in other software specialised for this).
Whether you have sufficient information to conduct the analysis is hard to tell. For example if you had 20 independent studies and a mean, n and SD for each one and were interested in the average effect size this would be sufficient information. If you also had moderator variables (i.e., knew the value of covariate of interest for each study) you could look at the effect of those on the average effect size.
- Badges
- Science topic
- Similar topics
- Medicine
- Epi
- Age Groups
- Children