Dear Gülizar Kaya ,

Your objective of study is not clear. Mention the objective(s) clearly and specifically. Also, clearly explain about the data you have for the study.

Dear mr Chakrabarty,

The objective of my study is to see if there is an effect of my IV on multiple (6) dependent variables. My dependent variables represent the BOLD activity of specific brain regions and together they represent a brain network.

My IV is a continuous variable and it appoints the spatial position of a brain tumor. I want to see how the position of a brain tumor affects global network activation changes in the brain.

Furthermore I have variables I want to control for.. For example, I expect that my IV has a different/stronger effect on my DV's if a tumor is in the left side of the brain as compared to a right-sided tumor. This 'hemisphere side' is a categorical variable, so what I expect is a significant interaction between this categorical variable and my continuous IV on my dependent variables.

I would like to test my hypothesis by putting all these variables in a model, while also taking into account that tumor characteristics, tumor volume, age, sex etc. could also have an effect on the DV's.

Dear Gülizar Kaya ,

You will have to test the significance of

(i) multiple regression coefficient

& (ii) partial regression coefficient.

For testing these two, you are to apply F test.

Go through an standard book of test of significance.

Dear Colleague

I strongly recommend that you use repeated measures design on the basis of mixed model methodology through mixed procedure of SAS package program. Please see repeated measures design papers in my researchgate page.

If you want to obtain a prediction equation for each dependent variable

I would like to invite you to use Multivariate adaptive regression splines as a powerful data mining algorithm. Other alternative is multivariate multiple regression analysis.

Warmest Regards

Prof. Dr. Ecevit EYDURAN

Please state you objectives clearly

Obvious Nchimunya Chilyabanyama Thank you for your comment. However, please see the reaction to Mr. Chakrabarty. What exactly is not clear enough? I would like to explain it better, but I don't know what is not clear to you.

Ecevit Eyduran Thank you for your reply. However, what is your argumentation to use a mixed repeated measures design above multivariate regression analysis?

Unfortunately, I do not have acces to SAS/STAT. I can use SPSS or R statistics (the latter I have no experience with, hence why I tried the GLM Multivariate procedure in SPSS).

Dear Gülizar Kaya,

Mixed model methodology involves various variance-covariance structures in repeated measures designs. The link regarding R may help you

http://www.jason-french.com/tutorials/repeatedmeasures.html

Best Regards

Structural equation modeling can handle multiple dependent variables simultaneously. Based on your research question, it seems like you don't need to create latent variables. If that's the case, it would be pretty straight forward to test the path analysis. In SPSS, you can explore AMOS add-on package. In R, you can use lavaan package.

Dear Concerned

It is rare though to have 6 DVs against 1 IV, however, if you use multiple regression in EXCEL it will give you the ANOVA Table which you use for values of p compared to level of significance, and F sign, compare also to level of significance for the entire model test. Thank you