It sounds like the issue may not be related to low N then. One other thing to check would be syntax. Occasionally one measure with bad syntax can mess up a table. Do you think that could be it? You could try deleting measures one by one to see if you can find an offending measure. – robsmith 4/26/2018 8:47:37 PM

I thought it was a syntax issue too. But I'm not writing it in manually. I'm selecting frequencies under available measures, and adding levels that way. – user-b432f7cd 4/26/2018 8:50:17 PM

Here is the entire table's syntax: https://ibb.co/hNdNYc – user-b432f7cd 4/26/2018 8:51:47 PM

(Edit: this comment is incorrect) Thanks for sending the syntax! Since you didn't write the syntax manually, I'm guessing you used the Reduce Base window and entered multiple values in the box. Only one value is supported. So that window created invalid syntax for you...

The nullifeq function only supports two arguments, with the second argument being the value to exclude from the base. In order to exclude multiple values from the base, you'll have to manually edit the syntax:

Change: nullifeq(r1,-7,-8,-9) to: nullifeq(nullifeq(nullifeq(r1,-7),-8),-9) – robsmith 4/26/2018 9:06:19 PM

Thanks, I'll try that. But I'm confused. Why would nullifeq fail only for counts, but work fine with percents? – user-b432f7cd 4/26/2018 9:33:01 PM

That didn't work for me either: https://ibb.co/cX2XYc – user-b432f7cd 4/26/2018 9:35:43 PM

Also, I tried using that syntax format in another data file with larger bases, and it worked perfectly fine. It's only going blank for bases under 50. – user-b432f7cd 4/26/2018 9:36:35 PM

Good point about it working with percents. After further investigation, I see that nullifeq actually does support multiple args. I think you've found a bug with the countin function. I'm investigating it further. – robsmith 4/26/2018 9:38:19 PM

I have a question... are you computing weighted frequencies? And if so, is there a chance that your weight has missing values for the records where r1 has a value of 1,2,3,4 or 5? There may be a slight difference between the way countin and pctin handle missing weight values. – robsmith 4/26/2018 10:02:09 PM

That was it! There was a weight var assigned but not actually present in the dataset. When I disabled the weight the frequencies appeared. There must be a difference in how pctin and countin interact with the weight var, because only countin is affected when a weight is assigned but the var not present in the dataset. – user-b432f7cd 4/27/2018 1:41:46 PM