StatsIQ - Different correlation results if I use text set or number set | XM Community
Skip to main content

Dear all,
I'm seeking to correlate different key drivers to a NPS.
The key drivers exist both in text set (not satisfied to fully satisfied) and in number set (1 for not satisfied and 5 for fully satisfied).
When I use regression analysis from StatsIQ with the NPS as the key variable, the results are different depending if I correlate to the text set or the number set. If I do the correlation using text set drivers, let's say that drivers #1 and #2 will be the most correlated to the NPS. But if I do the correlation using number set drivers, drivers #3 and #4 will be the most correlated to the NPS.
Thus, I'm not too sure which analysis is the one I should rely on to draw conclusions.
Have you already faced this question? Would you be able to share some insights on how to interpret the results?
Many thanks in advance.
FJ

Check your re-code values. I don't know if you use the automatically generated scales or not, but these typically have recode values that do NOT automatically move sequentially (1-5). If your recode values are different than the 1-5, it could be correlating to those.

Hi Kate,
Thanks a lot for your answer and for taking the time to solve my issue.
I did recode the value so I'm not sure the issue lie there but I might be wrong, I share with you the recoded values. Do you think it might explain my issue?
FJ
Capture.PNGCapture2.PNG

Hi FJO! If you haven’t already, we’d recommend reviewing the Q-Tip on this support page about Regressions to help you get started. If you still have questions after reviewing the linked support page, you'll want to reach out to the Support team, as they are best equipped to help answer questions and provide support! 😁

Leave a Reply


Terms & Conditions