In my study I have two samples which are independent. After finding significant differences between the groups on few variables, following intervention in one of the two research groups, I took simple Pearson correlations between pairs of variables in each of the study conditions (time, group), but as I have three variables it seems like I have to use partial correlations too. Especially when considering that Pearson's correlations between two of the variables are around 0.8 (which is in line with scientific literature about them). I'm talking about two creativity tests and another personality test, all are continuous variables.
I'm not that used to correlations as part of the study design, so I have few questions: Is using partial correlations should be the first step when there are more than two variables in the study design (with the prior assumption that they all correlate with each other)? Second, in such case should one avoid using Pearson correlations? If not and using Pearson correlations is also asked, in what order should it be done (meaning naturally I assume that first Pearson correlation and then partial correlations?) and if I include one table of correlations would it be this of the partial or the Pearson's correlation (or both)? Of note, in previous studies there was found no difference in the strength of the correlation between the two creativity measures and the personality measure and this is my assumption too.
Another question, after I find all correlations are significant, I used z tests for independent and dependent correlations (some of the comparisons were between groups, other were in the same group but before and after) using the online calculators from this site: http://www.psychometrica.de/correlation.html
How should I report the Z tests (I think it's modified Steiger's z tests), dependent and independent, in APA style (in text) considering it was taken from this website and that I used two different z tests (1 and 2 on the website). Also, I used Pearson r in these z tests, should I use partial correlations instead (at least for test 1, after all I have three variables, but in each comparios I use pair of correlations)?
Thanks to all
http://www.psychometrica.de/correlation.html
Dear Gilad,
here is the answer in this link:
http://www.real-statistics.com/correlation/multiple-correlation/
Dear Gilad,
here is the answer in this link:
http://www.real-statistics.com/correlation/multiple-correlation/
Beatrice is absolutely right, multiple regression/correlation analysis seems to be the method of interest for you. But checking for assumptions, conduction, and interpretation of the results may be very complicated and confusing and not as straight forward as a Pearson correlation. Therefore, I would recommend a good textbook to get into this field (e.g. Andy Field - Discovering Statistics Using SPSS [also as SAS and R version available]) or to contact a statistician in your department to avoid any pitfalls that may occur.
https://uk.sagepub.com/en-gb/eur/discovering-statistics-using-ibm-spss-statistics/book238032
Thank you both. Reviewing previous studies very similar to the one I'm currently working on, they used Pearson and partial correlations as well as z tests. This is pretty much my first study which is very correlational in nature, I agree that multiple correlations seems better way.
Partial correlation is a value you get from regression analysis to calculation what % a certain predictor predicts of the dependent variable.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis