My study is to investigate the effect of energy density (ED) and portion size (PS) on the intake of food among 12 – 16 years old female students. There are two levels of ED (High and Low) and PS (Large and reference) is manipulated in the provided test meals for lunch.
My 40 respondents had gone through four trials : (High ED, Large portion), (High ED, Reference portion), (Low ED, Large portion) and (Low ED, Reference portion).
Based on the journals that I read, they used Two-way factorial ANOVAs to test the effect of portion size, ED and their interaction on the test meal (weight and energy intake), other food (weight and energy intake) and meal energy.
As my objectives for my study area :
To test the effect of ED, portion size and combined effect (ED&PS) on the intake of
(a) test meal (weight and energy)
(b) lunch meal (weight and energy)
(c) subsequent meals (weight and energy)
I used 2-way ANOVA Repeated measures SPSS. Based on the journals, all models are adjusted for age, ethnicity, BMI z-score and test meal preference. I included 7 covariates in my analysis.
However, none of my results were significant which was totally different from the journals.
1. Has my analysis been wrong?
2. Should I use Univariate analysis?
3. As I use raw input (weight of the food intake), should I compute about my raw input to something before I analyzed?
40 is a tiny sample 4 conditions. You have not randomised to conditions and so ANOVA use is plain wrong – however many others have made the same mistake. The term ‘significance’ can have no meaning here. Ditch it.
Simply quote the ‘effect’ sizes or standardised R-square or whatever. If this shows no relationship then that is your finding. Accept it and use it. Don’t dredge for soemething else you want to find. That is not research. Anyway all of the other studies may be wrong. Most research is wrong after all.