Hi, new grad student here!
I have a limited understanding of statistics at the moment. I am writing my thesis proposal now and I am teaching myself the stats along the way. This is an ag bases experiement. I have 3 treatments (control and two treatments) and I'm measuring 3 soil properties within those treatments. There will be two soil data collections within the experiment on all 3 soil properties (pre and post).
My advisor has said that it will most likely be repeated measures ANOVA, but I'm having a hard time fitting my design into an RM ANOVA. I understand that I'm measuring the soil properties over time, but I'm only collecting data in two time periods and comparing those. RM requires at least two or more time points.
Any help would be much appreciated.
If the 3 soil properties you mentioned convey a single, general concept, then your study is One-way MANCOVA. Where each Pre is placed as a covariate.
If the three soil properties are each a separate concept, then your study is One-way ANCOVA, in which each Pre is placed as a covariate.
Of course, you are not saying that it is not RM study. See the link below.
https://www.theanalysisfactor.com/pre-post-data-repeated-measures/
or
https://graphpad.ir/repeated-measures-or-ancova/
Sometimes we use RM with only two times before and after
The important aspect of a repeated ANOVA is that you have repeated observations across variables i.e. at least 1 dependent variable has more than one observation. Also, advisors don't always have expert statistical knowledge, so you could find the person who does, either in your faculty or another :)
Given the complex nested characteristics of your variables, might I suggest using hierarchical regression (linear mixed effects model) instead of rANOVA? This will give you greater modeling flexibility, such as the ability to examine crossed and nested random effects - ultimately allowing you to examine more complex relationships among variables and obtain better estimates.
Abolfazl Ghoodjani The soil properties are separate concepts (bulk density, aggregation, compaction) so I'm thinking one-way ANCOVA. However, I don't really understand how the "pre" measurements can be considered covariates? From what I'm reading covariates are for describing either unwanted interactions or independent variables (the three treatments in my case).
Thanks for your response!
An additional question that I have:
It seems that RM ANOVAs break things down by subject. Can my subjects be my treatments? I'm trying to compare my 3 treatments against 3 dependent variables over two time periods. It looks like that might work for RM but I'm having a hard time seeing how it will fit.
Thanks, everyone!
ANOVA is only able to be conducted with a single DV. This situation requires a MANCOVA or multivariate regression.
Megan McCoun See the text below. Very well written. Comes with details.
By reading it, you will understand why and how you should use One-way ANCOVA in SPSS.
https://statistics.laerd.com/spss-tutorials/ancova-using-spss-statistics.php
Running ANOVA is your best bet, and you can fix this within your model. Since you are also measuring responses at different (pre and post), this is RM and with subject nested in the treatments.. Your subject cannot be your treatment in this case.
Megan McCoun A multivariate mixed-effects model will allow you to examine the effects of subject AND treatments in a single analysis. ANOVA/ANCOVA/MANCOVA cannot do this. They would require a separate test for subject and treatment. The responses above me do not seem to have read your description very carefully. An ANOVA/ANCOVA can only handle a single DV, but you have 3 DVs. This situation requires a multivariate test.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA