Hi

This depends of course on what your research problems are. If these involve predictions only, multiple regression would be fine. If you want to specify models with direct and indirect relations, structural equation modeling (amos, lisrel, mplus, lavan ...) would be a nice alternative. If you want to analyze moderation models, multiple regression with interaction effects may be used. This decission should start in your research problem.

Best, Rolf G

Just to be a wet blanket, where is your dissertation advisor? I would discuss things like that with all of my dissertation students. Best, David Booth

yes

David Eugene Booth sadly I was told to read papers and just figure it out

Rolf Gjestad thank you so much for your insight the aim of my research is to look at whether academic stress effects premenstrual symptoms in university students. But I'm also interested at looking at how social demographic factors: ethnicity, age and level of study can also effect premenstrual symptoms. It would be of interest to me to look at the possible interaction effects my IV's could have so I think I will take your advice and do an multiple regression with interaction effects. Just to confirm it but because my dependant variable scale is ordinal (likert type) so would it meet the assumptions for a multiple regression? because from what I was reading online a ordinal regression may be better- the only thing being I'd have to treat my stress scale, which is a likert scale too, as either categorical or continuous. If in that case I think I would treat it as categorical as I could group participants in not stressed/ moderately stressed/ highly stressed right?

Sorry for asking so many questions it's just a stressful time and I've had a lack of support around me.

Also Miqdad Jasim thank you as well.

I apologize for your boorish advisor. some of us are such that what you learn from us is what NOT to do when you have research students. Rolf Gjestad is on the right track but you Have Likert variables. I agree that what you do with statistics DEPENDS on your research question. However since you have likert variables and you wish to compare premenstrual symptoms with academic stress . The first with three likert scales and the second with two. The DVs will depend on your specific research questions but either way the statistical approach is the same. Lets choose the simplest to illustrate. Suppose you want to look at academic stress as a function of premenstrual symptoms. So you have F(DV1, DV 2) = f(IV1, IV2, IV3) F and f depend upon the considerations that Rolf Gjestad mentioned. Do you want to study interactions and so on. but this is not multiple regression in the ordinary sense. The RHS of your model will be the same as a multiple regression RHS with likert variables. The LHS will in this example with be a Multivariate regression with 2 DVs This is futher complicated by the fact that the DVs arer Likert and hence ordinal. This is not OLS. In the case of 1 DV this would be ordinal logistic regression. My fav reference is attached..

Now we just had some Likert Questions like this, see: https://www.researchgate.net/post/What_would_be_the_best_linear_regression_strategy and: https://www.researchgate.net/post/In_the_ordinal_regression_model_should_estimates_coefficient_slopes_and_thresholds_modified_to_accommodate_significant_explanatory_variables_only So look at ordinal logistic in the notes. Read the posts in the two Researchgate questions. Then Since whichever dependence you wish to study you have a multivariate ordinal logistic regression. The difference is the number of DVs. The way to handle these is to look multivariate ordinal logistic regression. I recommend the R package because you can find R examples of all of these things once you look at another link:

https://www.bing.com/search?FORM=U528DF&PC=U528&q=multivariate+Ordinal+logistic+regression+with+R

Whew. undoubtedly you will have questions. so be prepared to ask. to get help with the RHS of your models I suggest this link:

https://www.bing.com/search?q=Montgomery+D+C+Design+and+analysis+of+experiments&qs=n&form=QBRE&sp=-1&pq=montgomery+d+c+design+and+analysis+of+experiments&sc=1-49&sk=&cvid=952984F9DF6E4FF0A21CA0A1EE235A05

use the regression model form NOT ANOVA Form.

Now we come to the crux of your advisors problem. He had not a clue about any of this, I suspect. ALWAYS start with a research design and a plan for attacking this with appropriate statistical methods. I apologize for giving you even more reading but there it is. I believe that if you clean up your research questions a bit and work through this you will have quite a nice dissertation.Don't let your advisor talk you out of any of this. You have sold literature to back you up. Look this mess over and feel free to ask any questions that occur to you. Best wishes, David Booth.

David Eugene Booth I've only ever learned how to/ have access to SPSS. I've had a read of some of the links and I feel I understand ordinal logistic regression better. As I've looked over this article by Laerd statistics too: https://statistics.laerd.com/spss-tutorials/ordinal-regression-using-spss-statistics.php

I've had to change my measure of premenstrual stress and this new one doesn't have any sub-scales. I believe that due to this change that I will now have to use a ordinal logistic regression as opposed to a multivariate ordinal regression. Am I correct in making that assumption? Also I'm looking at the possible moderation effects my social demographic variables can have on the relationship between premenstrual symptoms and academic stress so I'm thinking a linear logistic regression with interaction effects is best for my work?

Zara Nadeem Without seeing things in detail, It appears your conclusion is correct. I would not recommend SPSS for your project. I'm sending some R stuff that I would like you to look over first to get started(including how to get the free download): http://sgpwe.izt.uam.mx/files/users/uami/gma/R_for_dummies.pdf

Examples in R for ordinal logistic regression can be found here, including tutorials:

https://www.bing.com/search?FORM=U528DF&PC=U528&q=Ordinal+logistic+regression+examples+R and attached the best notes I know of on the planet with examples in R. If I can be of assistance, feel free to ask. Best wishes, David Booth

Dear Zara Nadeem

I agree with you that you can use multiple regression with Step wise.

With my best wishes