How do I select the best distribution for analyzing a non-normal data in Glimmix?

[JismolJ]

I did a growth performance study in grower pigs using 4 different diets ( peas and enzymes were the two factors)and the data turned out to be non-normal. The experiment was arranged as an RCBD on factorial design. 

What distribution should I use...gamma or poisson?

How would I identify the fit statistic? Can I get AIC, BIC values using glimmix or should I use proc mixed?

How can I get the estimate values if I am using any of these distributions? Or should I take the estimate values from proc mixed?

This is the code I have used:

ods html;                                                                                                                     

ods graphics on;                                                                                                                       

proc glimmix data=Body_weight plot =(residualpanel studentpanel);                                                                                                                

class block pen peas enzymes Day;                                                                                                                   

model weight =  peas|enzymes|Day/dist=gamma link=log s;                                                                                                                

random Day/subject=Pen(peas enzymes) type=cs residual;                                                                                                                   

lsmeans peas|enzymes|Day/slice=Day;                                                                                                           

run;                                                                                                                

ods graphics off;                                                                                                                      

  ods html close;

I have attached the entire code with my data for further information.

Please guide me.

Thanks.

[PGStats · Accepted solution]

When fitting

proc glimmix data=Body_weight plot =(residualpanel studentpanel);  
class block pen peas enzymes Day;
model weight =  peas|enzymes|Day/ /*dist=gamma link=log s*/;
random Day/subject=Pen(peas enzymes) type=un residual;
lsmeans peas|enzymes|Day/slice=Day;
run;

the studentized residuals look like

I don't know about the model. But the residuals look fine to me.

[JismolJ]

Thank you so much.