The LOGSELECT Procedure
WEIGHT Statement
WEIGHT variable </ NORMALIZE>;
The variable in the WEIGHT statement is used as a weight to perform a weighted analysis of the data. Observations that have nonpositive or missing weights are not included in the analysis. The values of the WEIGHT variable can be nonintegral and are not truncated. When you omit the WEIGHT statement, all observations that are used in the analysis are assigned a weight of 1.
You can specify the following option in the WEIGHT statement after a slash (/):
- NORMALIZE
normalizes the weights that are specified by the WEIGHT variable so that they add up to the sample size. Weights are normalized by multiplying them by
, where training observation i has a frequency equal to 
and a weight equal to 
, and 
is the number of trials when you are using events/trials or multinomial-trial syntax; otherwise is equal to 1. When you specify this option, the estimated covariance matrix of the parameter estimators is invariant to the scale of the WEIGHT variable.
Caution: PROC LOGSELECT does not compute the proper variance estimators if you are analyzing survey data and you use the WEIGHT statement to specify the sampling weights. Instead, use the SURVEYLOGISTIC procedure, which is designed to perform the necessary, and correct, computations.