I have two studies within my thesis;

The first is cross-sectional, and tests a theoretical framework of leadership - it explores the relationships between factors within the environment, the behaviour of the leader and their cognitions. For this, I have performed mediation analyses using structural equation modelling.

The second study looks at the impacts of a leadership training intervention, specifically;

1) comparing pre, post and follow-up scores on four training outcomes using four separate repeated measures ANOVAs, *and;

2) exploring the factors that influence the transfer of training (i.e. the scores on the training outcomes when assessed 6 months after training). This is done by performing *four hierarchical multiple linear regressions; one for each of the training outcomes' and controlling for pre-training scores (but treating.post-training scores as a predictor, along with other predictors of training transfer such as the level of post-training support received).

I am now starting to question whether my decision to use ANOVAs and hierarchical multiple regressions will be criticised when I could have used SEM. *The first study has a larger sample size (N=357) compared to my second study (N=205), which is why I originally went down this route, but I wanted to see if there is any research that could help me to justify NOT using SEM?

Could the fact that I have run several ANOVAs (one for each training outcome) and several hierarchical regression models (one for each training outcome) lead to type I errors? Would the *ideal situation be to *somehow model these all in one go using SEM?

I have read that SEM would have allowed me to 'explore links between variables whilst controlling for all other aspects of the model', but I'm not entirely sure what it means by this?

Any help would be greatly appreciated!