I have some 1000 treatments in augmented design. However, when I run F test for differences among treatments, it is not significant. Pie chart of the treatments according to frequencies of different treatments, there is a wide range of differences among the treatments though.
I have got some 944 genotypes which include some 7 replicated checks.
How many degrees of freedom do you have for the error sum of squares? I suspect that is where the problem may be coming from.
Yes. I checked the data for normality. It was not normal. But the reason for that was that it had outliers. I removed the outliers but even then it didn't work. What transformation can we use to correct data for outliers? Any clue?
In fact I did log-transform the data, but that didn't work as well. I haven't ever used non-parametric test though.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis