Pairwise exclusions I am running a logistic model on insurance data. I have a field agent gender which matters for a channel A (say) and doesn't matter for B. I want to put null values in case of B. The only thing I risk is exclusion by SAS (as SAS excludes every missing case by default). I heard that pairwise exclusion can solve my problem. Please tell me the following:


*

*What is the difference between normal exclusion and pairwise exclusion?

*How to do this is SAS?

*What is listwise exclusion?
 A: This doesn't sound like a missing data problem to me: it sounds like a question of model structure.
Distilling it to its essence, it seems you have two independent categorical variables gender ($X$, say) and "channel" ($Y$) and a binary response ($Z$).  Conceptually the model is
$$logit(\Pr(Z=1)) = \beta_0 + \beta_1 X + \beta_2 Y + \varepsilon$$
when $Y$ = "A" and otherwise
$$logit(\Pr(Z=1)) = \beta_0 + \beta_2 Y + \varepsilon$$
when $Y \ne $ "A" (with parameters $\beta_0$, $\beta_1$, and $\beta_2$ and zero-mean random error $\varepsilon$).  If this interpretation is correct, just set $X = 0$ in the data whenever $Y \ne $ "A".  This will cause such cases to have no effect on the value of $\beta_1$, which will be estimated solely from the other cases where $Y$ = "A".
A: There are two main types of traditional treatments of missing data.
These are:
1) listwise
2) pairwise
Listwise is (from what you have said) the default in SAS. It means that you exclude any observation that has missing values in any of the terms in your model.
The advantage of this is that it ensures that all variables have the same n in the model. The disadvantage is that if you have one variable with a high proportion of missing data, much of the rest of the data can be ignored. 
Pairwise, by contrast, only excludes cases where the value is missing for that particular variable. This has the advantage of keeping more data in the model that listwise does. However, the disadvantage is that some of your results will be based on different subsets of the data, and this can cause problems with p values and confidence intervals. 
A better approach is probably multiple imputation where you attempt to predict the missing values using all of the data you have. This normally involves simulating a number of datasets with the missing data filled in, and then performing all analyses on each of the datasets individually, and averaging the results.
A good paper on treatment of missing data is Graham 2009 Missing Data: Making it work in the real world which goes into much more detail than I have here. You also need to determine if the missingness in your data is random or not, as if the probability of missingness depends on exogenous variables to your dataset, then none of these approaches will work correctly. 
A good resource on multiple imputation is here
This article on missing data in SAS may prove helpful, or someone else may answer as to how to do this in SAS. 
