Use both season and month in model?

[always-good]

Dear all,

I have an example for which I'm confused to set the best model.

I want to see the effect of the season on the milk production of crossbred cows. Seasons are: Summer season= Jun-Jull-Aug Winter season= Dec-Janv-Fev; There are two groups : 40 animals with the same lactation stage and the same characteristics in each season.

can we say: Yij=u+Si+ei (GLM);  OR Yijk= u+Si+Mj+(SxM)ij+ek, where m = random effect of the month

but I think it is useless because the month is included in the season. Which is the best model or is there other options.

Looking forward to an answer

[PaigeMiller]

Since month is included in season, I would not put both in the model. Just use month. You can create a contrast that compares the effect of June/July/Aug to Dec/Jan/Feb.

[always-good]

the main effect is basically the season, can we say that month is a repeated measurement? in this case is this a mixed linear model? and how to present the month in the result table please?

[PaigeMiller]

---

@always-good wrote:

the main effect is basically the season, can we say that month is a repeated measurement?

---

As you are the subject matter expert, and not me, these are decisions you need to make. I have no opinion one way or the other as I don't work with cows or any type of similar study.

in this case is this a mixed linear model? 

If you decide these are repeated measurements, then I think so.

and how to present the month in the result table please?

what table are you talking about?

[always-good]

Hello, 

Sorry for the misunderstanding, I'm asking if both options are correct first to decide whether it is GLM or can be mixed.

I mean to present the effect of the month as a repeated measurement. 

Thank you 

[Ksharp]

" I think it is useless because the month is included in the season. "
It is depended on what you want,
if you only focus on season effect,you don't need month effect any more.
if you also take into account of month effect within season effect,you could take month effect too.

If you think each cow have different milk production in different season,You can use MIXED model.
If you think all the cows in the same group have same milk production in different season,You can use GLM model.

In summary,I would use Mixed model if you take a cow as a SUBJECTID.
if you only consider the population of cow within a group I would use GLM .

And better post it at Stat Forum:
https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures

@StatDave @lvm @SteveDenham could give you a hand.

[mkeintz]

If you know the month, you already know the season.

When one predictor is entirely predictable from another predictor, I believe that the resulting SSCP matrix will not be invertible, as it won't be full rank.