Dear colleges,
I was conducted two field experiments in two consecutive years in order to estimate genetic diversity analysis among forty fourth genotypes of lentil (lens culinaris M.). Morphological characterization was done based on different parameters like plant height, pod per plant, seed per plant, 100 seeds weight… The experiments were carried out following an Alpha Lattice design with fourth replicates. Data were recorded from both years. I would like to know if generalized linear mixed model (GLMM) is applicable in this case, considering a genotype as a fixed effect and a bloc as random effect.
Thank you for your response.
Best Regards,
You may follow the original paper for doing an LATANOVA http://libcatalog.cimmyt.org/download/agris/62983.pdf, but the original package is in MSDOS. R-project has got support for alpha lattices in Agricolae of Dr Felipe de Mendiburu specific pachage. R-project is a free stats system for PC. https://cran.r-project.org/web/packages/agricolae/vignettes/tutorial.pdf. The variance covariance of some trait can be a further object of analysis but it is possibly a secondary aim if features are related in some way in a model, https://cran.r-project.org/web/packages/glmm/vignettes/intro.pdf
GLMM will be aplicable if you have a non-normal response variable (like survival), but it seems you have several continuous variables (such as height), hence, you are fine using LMM (that is acutally equivalent to GLMM with normal response). This will be the appropriate form, and there are many packages in R and SAS that can handle mixed models. The issue will be if you have a pedigree, that needs to be incorporated. In this case use ASReml-R that will handle this nicely. However, you are considering your genotype as a fixed effect so no need.
Now, as you have an Alpha-Lattice, the key point here is that you need to consider block random otherwise your analysis will be incorrect.
Dear Djouher,
I agree with Salvador. If using GLMM, that would be more robust since the data may not be normally distributed. Actually, I would not suggest using ANOVA in any situation, since ANOVA requires too many strong assumptions, such as homoskedesticity, which is very hard to verify. In my opinion, ANOVA's only advantage is that it is easier to be computed by hands, but in this era, who needs to do empirical studies by hands?
Hope it would be any help to you.
Sincerely,
Kuan-wei
Hi, If you are after a decision whether to use GLMM. This is one of the best readings. http://www.cell.com/trends/ecology-evolution/pdf/S0169-5347(09)00019-6.pdf
Hi,
I would like to thank’s all of you for your assistance and help. Actually, I am using GenStat software.
I do normality test for all my data, but it’s seems they don’t fit normal distribution even after transformation.
I will be grateful for much more suggestions.
Thank you,
Djouher
Dear Djouher
Salvador gave you right comment , and in the case of unbalance data ( may be due to different survival ), you must use GLMM.
Also, you need to test your data for normal distribution .
Good Luck
Thank's Dr.Khalid.
Another questions is about categrical Data like scal from 1 to 9;
Wich analyziz is suited in this case?
Thank's
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA