I guess you mean an automated way. I don't think there is an automated way for this. But if you are sure to delete then you should be able to use well known methods from within PSPP.

I recommend against using any automated or blind way to identify or remove "outliers". What you are tempted to call an "outlier" may just be a legitimate and valuable observation... At an initial glance, PSPP doesn't give you too many options here. If you are assuming a variable is normally distributed, one thing you can do is Analyze >> Descriptive Statistics >> Descriptives >> Option: save z-scores as a separate variable. And then you can identify observations with a z-score greater than say, 2 or 3, (less than -2 or -3). ... It also appears that you can run the model you want and extract the residuals ( https://underthecurve.github.io/jekyll/update/2016/07/01/one-regression-six-ways.html#PSPP ). (But I didn't try this.) Examining the residuals may also be useful to identify observations that could use a second look.

Thank you so much for your valuable insights.