Hello Ella,

Generally, it's easier to sort out how to proceed with data analysis if the specific research question/s are identified in advance of data collection, rather than after the fact. I suppose the simplest way to go here is to ask, what would you like to know from these data? (The follow-up questions would concern what variables you have, how they are quantified, and data collection.)

From your query, it sounds as if multiple regression, path analysis/SEM, or multivariate regression are all possible analytic approaches. I'm sure that your institution has a number of folks with whom you could sort out your aims, the nature of the data you have, how the sample was drawn, and other relevant features of your study to recommend an approach that would be satisfactory and defensible.

Good luck with your work!

Dear Ella Malloy ,

As you only have two discrete time points you could, provided your DVs are scale you could put everything into a MANOVA. However, as David Morse rightly says, you need to be guided by your research question. I would always recommend trying the simplest model with the maximum potential explanatory power (a principle known as Occam's razor).

It depends on the type of variable that you are studying, paired t test for quantitative data, repeated measure anova if u measured more than two times.

In case of categorical data, macNemar test for categorical variables of two levels and measured twice, Cochran if measured more than two times,

Hello David Morse

Thank you for your answer. The aim of the study was to see how a certain style of leadership related to changes in the outcome variables of athletes over time i.e. between time 1 and 2. So essential a direct relationship but also whether the leadership variable would indirectly effect the outcomes via causing changes in variables such as trust and team climate, over time.

The data was quantitate data collected via the same questionnaire scales at time 1 and time 2.

I wanted to check whether these were appropriate questions for the type of data before proceeding with the analysis? I have analysed the data using OLS regressions but my supervisor was not sure that is the best option, would SEM be better?

Thank you

Hello again, Ella,

It might be easier to set the model up as an SEM for analysis. I still can't tell whether the variables you mention are manifest or latent; that distinction would obviously influence the complexity of the model. No matter; you can certainly evaluate the impact of leadership style as a moderator, mediator, or just plain messy variable however you see fit within the SEM framework.

Good luck with your work.