I have model where I have four dependent variables in nomological order and 3 independent continuous variables . To test casual relationships I first compared means of outcome variables across different groups (based on median split in form high and low) of IVs through ANOVA. Than I tested the overall model and calculate mediation effects by using SEM through AMOS. Can i report the results acquire from both methods or i have to stick only with SEM?
SEM and ANOVA are two different issues. You can see the mediators relation in SEM in an integrated approach that's not easy to do in other regression or ANOVA.
SEM is just an advanced path analysis and we can find the direct and indirect relationships of latent variables together in a single frame.
Thank You Sir Muhammed Ashraful Alam .
I actually did both. first i convert my independent variables into categorical and conduct ANOVA for comparing mean scores because my dependent variables were measured on only one item. Than I performed an CFA trough AMOS on independent variables and test the whole model. I am actually confused whether to report ANOVA results or SEM results or can i report both because i have different results in both cases.
SEM is far better than normal regression or ANOVA. Significance is not the only issue.
No. You cannot. I will shortly talk about the process.
Hypothesis: There will be a significant difference in body fat among the group of militaries, social workers, and advocates, based on the notion that they do different works.
You do: ANOVA
Hypothesis: There shall be body fat 3-4 times higher in militaries than advocates, but there shall be 2-3 times lower fat in social workers.
You do: Regression
Hypothesis: All the 40 responses on body measurements given by the militaries, advocates, and social workers will eventually indicate the degree of fatness among them.
You do: SEM.
Hi Muhammad,
Short answer: NO.
1) Recoding continuous predictors to binary variables is a bad idea. You are losing information.
2) If your research hypothesis is about testing a mediation effect, then you cannot use ANOVA.
My suggestion: Only provide the results based on your mediation SEM model.
Best,
Roberto
SEM in this instance provides a more detailed description of the results. While it doesn’t encompass the findings of the ANOVA, the ANOVA results are still redundant after the SEM.
William Beckwith Thank you for your answer. Is it possible to report my analysis to divide in the form of two different studies. first study presents the results of ANOVA and 2nd study provides the major hypothesis testing based on SEM.
I forget to mention in my question that i have an experimental design in which i manipulate one variable through vignates. in experiment research design researchers normally adopt variance analysis in the form of ANOVA t test & MANOVA.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Modeling
- Analyses of Variance