topic Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec in SAS Programming

Proc Glm, Slice command with one class variable and three continuous variables - contrast effect

Lena_PA — Tue, 07 Nov 2023 07:14:19 GMT

Hello

I would like to test the variable "beta", "gamma", and "alpha" (these are continuous variables) in two different level of "pet_type". Pet_type is a class variable with two values: "dog" or "cat". Ultimately, I would like to have a graph something like this

my code is

proc glm data = have ;
class pet_type ;
model logsale = beta gamma alpha pet_type pet_type*beta pet_type*gamma pet_type*alpha/ss3;
lsmeans pet_type*gamma/ slice = pet_type;
lsmeans pet_type*beta/ slice = pet_type;
lsmeans pet_type*alpha/ slice = pet_type;
run;

but I've got an error message that "only CLASS variables allowed in this effect." Please help!

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Ksharp — Tue, 07 Nov 2023 07:34:58 GMT

Since LSMEANS compare the mean of Y in two or more groups within CLASS variable, like ANOVA . You should use CATEGORY variable in it.
lsmeans pet_type*gamma/ slice = pet_type;
-->
lsmeans pet_type/ diff ;

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Lena_PA — Sat, 11 Nov 2023 03:51:33 GMT

Thank you for letting me know! Pet_type is a category variable but alpha, beta, and gamma are not.. so what do you suggest?

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Ksharp — Sat, 11 Nov 2023 09:46:11 GMT

If You want to check the difference of ALPHA within a category variable,you should check this :

https://support.sas.com/kb/38/384.html

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Lena_PA — Thu, 03 Oct 2024 02:04:56 GMT

thank you! I reviewed the document and used the last one.

Pet_type is a class variable (either "cat" or "dog")

proc glm data=pet;
      ods select ParameterEstimates;
      class pet_type;
      model Y = pet_type alpha pet_type*alpha / solution;
      run;

The results show some value for cat but only zero and "." for dog. How can I get the results for both cat and dog?

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Ksharp — Thu, 03 Oct 2024 03:20:13 GMT

That was supposed to be , because 'dog' is reference level of class variable.

You could see X*DRUG G is also zero in the following picture(the parameter of reference level is supposed to be zero) .

Here " some value for cat " represent the difference of slope between CAT and DOG,check yellow part in picture.

X(or alpha) represent then solpe of DOG(reference level), check blue part in picture.

"get the results for both cat and dog? "

What do you mean by "results" ?

1) If you refer to mean of Y, "Intercept" is the mean of Y for DOG when alpha=0 .

"CAT" is the difference in the mean of Y between CAT and DOG when alpha=0, then the mean of Y for CAT=CAT+Intercept.

2)If you refer to slope of Y, "Alpha" is the slope for DOG.

alpha*cat is the difference in the slopes of the CAT and DOG, then then slope for CAT=alpha*cat +alpha .

Above all these you can find at the link I posted, just read it carefully and understand it .

Re: Proc Glm, Slice command with one class variable and three continuous variables - contrast effec

Lena_PA — Thu, 03 Oct 2024 19:37:28 GMT

thank you so much. I ended up manually calculated intercepts and slopes and draw lines in excel not the most sophisticated way but still worked