Lapis Lazuli | Level 10

proc model dummies

I am runnig proc model and when I add dummies in my regression the model converges but in the output T-values are B.

what does it mean and How I can solve that????

1 ACCEPTED SOLUTION

Accepted Solutions
Lapis Lazuli | Level 10

Re: proc model dummies

thanks I find the solution. just eliminate the constant in the model.

13 REPLIES 13
Super User

Re: proc model dummies

You likely over parameterized your model. How did you create your dummy variables? If you had a variable with, say 3 levels, did you include 2 or 3 dummy variables?

Lapis Lazuli | Level 10

Re: proc model dummies

this my model

disGDFF=a+a1*logage+a2*logTNA+a3*logMT+a4*EXP_RATIO+a5*turn+a6*logtnasq+a7*AG+a8*EI+a9*G+a10*LTG+a11*GI+a12*MC+a13*MRC+a14*SC+a15*MCG;

AG EI G LTG GI MC MRC SC MCG are style dummies variables. by line dummy will be 1 for one style for example G and 0 for others styles

Super User

Re: proc model dummies

@sasphd wrote:

this my model

disGDFF=a+a1*logage+a2*logTNA+a3*logMT+a4*EXP_RATIO+a5*turn+a6*logtnasq+a7*AG+a8*EI+a9*G+a10*LTG+a11*GI+a12*MC+a13*MRC+a14*SC+a15*MCG;

AG EI G LTG GI MC MRC SC MCG are style dummies variables. by line dummy will be 1 for one style for example G and 0 for others styles

Lapis Lazuli | Level 10

Re: proc model dummies

so I did not understand your question ?? can you please explain what did you mean by levels
How did you create your dummy variables?
if the style = AG so AG=1 else AG=0
Super User

Re: proc model dummies

If you have a categorical variable, say Sex, it has two levels, F and M.

Using whatever method you used to create the dummy variables did you create one or two dummy variables?

@sasphd wrote:
so I did not understand your question ?? can you please explain what did you mean by levels
How did you create your dummy variables?
if the style = AG so AG=1 else AG=0

Lapis Lazuli | Level 10

Re: proc model dummies

for each dummy variable is 1 or 0

and I have nine dummies variables

Super User

Re: proc model dummies

I can't see your data or code so I'm trying to make a generic example that we can both refer to, which is why I'm using SEX as the example.

If you had SASHELP.CLASS for example and wanted to include Sex in the model, which has 2 levels, F and M.

You would create N-1 dummy variables, where N is the number of levels.

So this means 1 dummy variable, such as Sex: 1 = Female, 0 = male.

I don't know what you did or how you created your variables (I have asked but you never answered the questions) so I'm assuming you created two dummy variables. Because of this, the dummies are linear combinations of each other so that there is redundant information. In this case, there cannot be an estimate for one of the dummy variables.

Good Luck.

Lapis Lazuli | Level 10

Re: proc model dummies

Thanks you.
Just to give you an idea of my data

id EI G LTG GI MC MRC SC MCG
1 0 0 0 0 0 0 0 1
2 1 0 0 0 0 0 0 0
Super User

Re: proc model dummies

You need to answer the initial questions.

How many categories did you originally have before creating dummy variables?

How many dummy variables did you created?

Lapis Lazuli | Level 10

Re: proc model dummies

How many categories did you originally have before creating dummy variables?

How many dummy variables did you created?

Super User

Re: proc model dummies

@sasphd wrote:

How many categories did you originally have before creating dummy variables?

How many dummy variables did you created?

You need 8, not 9. Because you can figure out the 9th from the other 8.

So you either get one that's reported as B, or exclude it from the analysis in the first place. This is known as your reference group, what all the other levels are compared to as the 'baseline' for example.

Super User

Re: proc model dummies

Lapis Lazuli | Level 10

Re: proc model dummies

thanks I find the solution. just eliminate the constant in the model.

Discussion stats
• 13 replies
• 939 views
• 0 likes
• 2 in conversation