I have the data of maize drought trials. There are 3 locations and 5 environmental conditions where the trails were done. Now, the problem is that the data are very unbalanced. There are different number of replications in each locations e.g. one location has 3 replications and another has 2 replications. Moreover, the number of traits measured in all the locations also differs. In one location I have 20 traits and in the other I have 15. I want to analyse and compare the performance of the genotypes across drought stress environments, non-stress environments and a combination of both. More precisely a combine ANOVA. So, is it possible to do so? I am asking because the data is unbalanced. So, can I take the mean values for each genotype from 2-3 replications and combine those for the ANOVA? Does this analysis make sense?
Thank you.
I guess that the variation in replicates with 2 observation samples makes these treatments sensitive to random effects but you can gain gain confidence with the R-package using a free tutorial and apply some of cases described under ea1 in https://cran.r-project.org/web/packages/easyanova/easyanova.pdf
Individual location analysis has no proble. Howeve, in order to do combined ANOVA, you first check the error variance for homogeniety. Secondly, you can drop the addtional data you have and focus on the 15 traits that you have from both locations and do the combined ANOVA if the error term is homogenouse. Third, you can use the SAS statement for unbalanced data set and move forward.
Thank you for the answer. I am using R for the analysis. Can you tell me how can I perform that in R instead of SAS?
There is a huge risk of doing a combined analysis with your experimental structure. It does not fulfill all statistical assumptions especially homogeneity of variance. Using an imbalanced replicated structure will reduce the model strength and increase the experimental error. I would strongly recommend analyzing the data individually for each location. You can present them together as separate results and consider doing a repeated measures for equality of means, if needed.
For R you may down load the r-project basic according to ypur OS from their CRAN site and follow a tutorial. Thre are complete informations with datasets in each package, like easyanova and tests for homogeneity of variance are a standard feature main core, https://www.r-project.org/, and http://www.cookbook-r.com/Statistical_analysis/Homogeneity_of_variance/
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA