For my work i often build regression models and usually I collect all my variables into a table in SQL first and then upload into SAS to do the building.
The table usually has 600+ variables and a target variable and an observation number 1-n for each row. It would be handy to be able to reduce these variables before the importing.
The best way to do this I think would be to compare each variable with the target variable and rank them so that I can pick the top 100 say for the next step. However to do sel corr(variable_1, target_variable) for all 600+ variables would be rather silly and i don't want to have to create a table with 600+ variables with 600+ correlations. What would be the most efficient way to do this?