DATA WITH ZERO AS DATA

[aska_ujita]

Hello there!

I have a data of residual milk of cows and in this data have a lot of zeros, meaning that the cow haven't any milk retained in the udder.

I am using PROC GLM to analyse milk production and residual milk, as recommended.

But in this specific data, the model (too high) and R square (too low) is not good at all because the number of information with zero as observed result.

I can't delete this information, aren't outliers...

How I can manege that??

Thank you!

Aska.

[Ksharp]

Try to use distribution GAMMA or TWEETIE to fit model ,better use TWEETIE distribution.

[aska_ujita]

Hello Ksharp, thank you for your help.

My procedure is like this, how can I fit the distribution?

PROC GLM (or should be mixed?); 
CLASS treatment po dl cow;
MODEL residual= treatment po dl treatment*po treatment*dl;
lsmeans treatment//pdiff stderr lines;
lsmeans treatment*op1/pdiff stderr lines;
RUN;

I have repeated measures of each cow by day of lactation (dl)

po is the parturition order (primiparous or multiparous)

Thank you.

Aska.

[Ksharp]

I am not sure.

PROC GENMOD; 
CLASS treatment po dl cow;
MODEL residual= treatment po dl treatment*po treatment*dl  /dist=gamma    ;   /*or  dist=tweetie*/