Logical tests based on values recoded to . ( missing) fail

[#1]

I tried a demonstration for a class yesterday that bombed badly.  I have the following data set on opinions about national and state government.

I used Recode to set the 3's (Don't Know) in Column 3 to . (missing). I did the same to column  5.  Then I used a formula to create Column 2, 

Either I am missing a logical step, or there is some sort of distinction between a missing value received in the data set and one created by recoding.  In any event, the numbers in column 2, starting at row 19 should be: 0 0 0 1 1 0 0 1 1 0  (through line 28).

My goal is to create column 2 as a filter, so I can partition the data set between records which have no missing or DKs on the target columns and those which have one or more missing or DKs.

The ultimate goal is to do a simple Y on X presentation with measures of association in which the table will be free of DK or missing responses.  I could not find a way to shut down either missing or DK responses in the analysis  by hiding them.

[#2 · Accepted solution]

What version of JMP?

The pictures you posted seem like something isn't working correctly. I'm suspicious of the ... vs ...2 names being confused somewhere, either in the formula (check the column 2 formula one more time, make sure it shows the ...2), or in JMP. Try renaming the two recoded columns to "a" and "b" (for example) and then check the column 2 formula again; it should show "a" and "b" for the columns it references. (Formulas in columns should automatically update to show renamed columns. If the formula doesn't show the new name, then ... can you carve out just a few rows and 5 columns that show the issue and send it to tech support?)

[#3]

Thanks, Craig.

I will send a corner of the data to Tech Support.  Renaming the columns does not solve the logic problem--which still could be mine alone.

Larry