Spotfire Ideas Portal
Don't think this is currently possible. Currently, to do a random sample, I create a calculated column for random value like Rand(1). Then I filter the data to my sample size (e.g. I filter out 0.0 to 0.7 if I want a random sample that's 30% of the original). Then I have to export the data and re-import it to the analysis as a new data set. Would prefer a simple "random sampling" option as transformation. Maybe have it use different sampling methods (e.g. stratified sampling) too.
It would be useful to have other sampling methods such as stratified, cluster, and multistage. These might perhaps be better suited as Expression functions since there are input variables... not sure.
If used as a Transformation, however, I imagine the UI could be similar to how the Normalization Transformation is today. Since filtering is required for this, it might have to be a Transformation and not Expression.
Hey Thomas, I made a video on the row filter method: https://youtu.be/pL0p0ivjBH8
Would still love to have this simpler though, if it's an easy enhancement. I use this often.
@Thomas - Yes, I have tried it with those two features and it works. Was looking for a simpler way if it's possible.
Would be useful for model development but potentially also in dashboard design. (i.e. design dashboard with a smaller data set during development, then delete the transformation before publishing to production).
Have you tried the rand method in combination with a row filter transformation?
Great idea and very valuable in model development!