A more detailed sample should illustrate this problem better. It is not clear what you desire to do for those zeros. Whatever, a new verison of SAS allows you to use zeros options on dealing with zero frequency or weights.

2# jingju11
我只是做个回归分析，用距离做权重，大于某一距离的阈值记为0，所以这样weights变量中才会有0值，但是其它变量里面都没有零值的。好像有零值的记录全被剔除在外，并没有进入回归模型。现在不知道该怎么办了。高版本的SAS是否可以解决这个问题？

3# WEICHEN360

I misunderstood your problem somehow.
In fact zeros option is only working on proc FREQ.
For either proc REG or GLM will exclude those observations with weight as 0 from the analysis.
When you assign a low value of weight to a record, that will contribute very less to the (weigthed) residual sum of squares, and that is also how weight works.