topic Re: Model is not full rank in Statistical Procedures

Model is not full rank

Epidemiologist — Thu, 16 Nov 2017 22:03:07 GMT

Hello,

after programming a proc reg statement in SAS, there is written in the output:

Note: Model is not full rank. Least-squares solutions for the parameters are not unique. Some statistics will be misleading. A reported DF of 0 or B means that the estimate is biased.

Note: The following parameters have been set to 0, since the variables are a linear combination of other variables as shown.

Some of my estimates are signed with DF B. Can I trust them anyway? If not, how can I solve this problem?

Are the two notes Independent of one another or exists a connection?

Thank you very much for your help.

Re: Model is not full rank

Reeza — Thu, 16 Nov 2017 22:07:02 GMT

Are you using categorical variables where you made dummy variables? Did you remember to use N-1 dummy variables?

Re: Model is not full rank

Epidemiologist — Thu, 16 Nov 2017 22:26:04 GMT

Yes I use categorical variables (exposures and covariates for adjustment) and I programmed dummy variables for them. What does N-1 mean?

Re: Model is not full rank

Reeza — Thu, 16 Nov 2017 22:41:05 GMT

When you created your dummy variables, if you have N levels, then you have N-1 dummy variables. Is that what you did?

Re: Model is not full rank

Epidemiologist — Thu, 16 Nov 2017 23:19:23 GMT

Yes I did it in this way.

Re: Model is not full rank

Reeza — Thu, 16 Nov 2017 23:25:00 GMT

So which variables are showing the issue? Its not a categorical variable? Are they linear combinations?

Re: Model is not full rank

Epidemiologist — Thu, 16 Nov 2017 23:43:46 GMT

The Problems (DF: B) originate from categorical variables only that are divided in dummy variables, mostly from 1 to 4.