Results from the CONTRAST, ESTIMATE, or LSMEANS statement may appear as Non-est indicating the quantity is nonestimable. Following are the most common reasons for nonestimability:

1. Not all treatment combinations are present.
2. Either effects are omitted that should be in the specified contrast, or inappropriate coefficients are used in the contrast.
3. Sufficiently precise numerical coefficients are not provided. Fractions like 1/3 should be specified to at least six decimal places.
4. There is a confounding problem that is associated with the data and the model.
5. There is no estimate for error in the model.
6. The specified contrast is truly nonestimable.

For CONTRAST and ESTIMATE statements, the major causes of the nonestimability message are reasons 1, 2, and 3. For LSMEANS statements, the major causes are reasons 1, 4, 5, and 6. Note that the LSMEANS statement specifies a contrast of model effects as detailed in "Construction of Least-Squares Means" in the Details section of the GLM documentation. For details on estimability of LS-means, see Cai (2014).

For more details and examples, see Examples of Writing CONTRAST and ESTIMATE Statements (SAS Note 24447) and SAS for Linear Models, Fourth Edition.