So, I started with a model with 5 IVs and 7 DVs. This model had a poor fit (RMSEA = 0.405, CFI=0.760 or so). Most of my hypotheses were not confirmed, i.e. many regression coefficients were not significant, so there were no relationships between many variables.
Next, I started to prune the model, eliminating the paths with non-significant coefficients, and remove two variables that seemed useless (one IV and one DV). Finally, the re-specified model was good (RMSEA = 0.042, CFI=0.982). Also, all the coefficients were significant.
My question is: how to report the results in my paper? Should I present both models and say "model 1 was not good so I modified it to model 2"? Or should I report only the second model and say something like "out of my hypotheses, only this and this are supported"? What are the standards and the common practices in the scientific world? Any suggestion would help.
(N.B. I am aware that I have used SEM in an exploratory manner, though it is a confirmatory technique, but I had to do that. I had a hard time collecting the data and I simply couldn't settle for the initial poor model.)
Thank you!
Next, I started to prune the model, eliminating the paths with non-significant coefficients, and remove two variables that seemed useless (one IV and one DV). Finally, the re-specified model was good (RMSEA = 0.042, CFI=0.982). Also, all the coefficients were significant.
My question is: how to report the results in my paper? Should I present both models and say "model 1 was not good so I modified it to model 2"? Or should I report only the second model and say something like "out of my hypotheses, only this and this are supported"? What are the standards and the common practices in the scientific world? Any suggestion would help.
(N.B. I am aware that I have used SEM in an exploratory manner, though it is a confirmatory technique, but I had to do that. I had a hard time collecting the data and I simply couldn't settle for the initial poor model.)
Thank you!