Why "non-est" in an lsmeans tables

[Beatrice]

I have 2 missing values. when i have no missing values i dont have that problem.

how can i resolve it? 

proc mixed data=data;

class a b bloc

model y=a|b

random bloc bloc*b;

lsmeans a|b;

run;

thank you!

[PaigeMiller]

One of the most common reasons is that you don't have any data in one of the cells of the interaction. 

[Education]

@PaigeMiller, so is it good to leave out the interactions and just run the fixed effects only?

[PaigeMiller]

---

@Education wrote:
@PaigeMiller, so is it good to leave out the interactions and just run the fixed effects only?

---

This is such a vague question, divorced from any real application or data, that it is impossible to answer. 

Sometimes you should leave the interaction in the model, and other times you should remove the interaction. It depends on a lot of things.

If you want a more specific answer, provide details of what you are doing.

[Education]

I conducted a seasonal experiment over a period of 4 seasons from animals of different age, where within each season data was collected once a day for 5 consecutive days (so my fixed effects are: season (4 levels), age (4 levels) and days (5 levels). Sometimes some animals worked and sometimes they did not. when i run a model y= season|age|days  using the GLMM the lsmeans of some FE are non-est, but when i reduce the interactions everything is fine. Can you help please.

Thank you.

Education

[PaigeMiller]

If there are some cells where there is no data, then you cannot really estimate interactions.

Is that the situation you have?

[Education]

Yes, thats exactly my situation and i do not feel like it is a good idea to drop some obs in order to remain with equal or no missing obs in some cells. There should be a model to run rather complicated to my knowledge for now. i think i need to get a book or do you have any articles that you can refer me to?

Regards,

Education

[PaigeMiller]

I am not aware of a specific book or article that deals with what to do in the presence of a cell with zero data.

i do not feel like it is a good idea to drop some obs in order to remain with equal or no missing obs in some cells.

As far as I know, those are your choices. It isn't really a matter of it being a "good idea" or not, it is a limitation of your data.