I have a repeated measurment of C fluxes data, the raw data have large variation either for the treatment or for the control group. So I firstly resampled the data by using bootstrap for 1000 time and get a new sequence of data for both treatment and control group. If I want to test the signficance of treatment effect, is it statistically solid to do the ANOVA based on data from bootstraping?
Why did you select ANOVA instead of a multilinear modeling approach? This way you don't have to rely on the assumption of sphericity, and you can use the bootstrap for the confidence intervals of the b-values.
(or you can use a robust repeated-measures ANOVA as described by Wilcox 2005 and inplemented in the WRS or WRS2 package in R)
I also would like to suggest to run more bootstraps if your computer allows you to (e.g. 2,000 or 10,000).
You are welcome.
For further reading I can highly recommend this book by Field/Miles/Field (ISBN: 9781446200469).
Good luck.
ofcourse we can use the ANOVA for bootstrap samples but robust design may be acceptable.
What did you bootstrap from the original data. Means, medians, to get the 1000 new data to run ANOVAs?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA