Incorrect JASP output
hi all, I was wondering if any could explain my obviously incorrect JASP output. So, for instance, the maximum should be 9, but the first table indicates it's 7. Same goes for the minimum, which should be 3, but JASP indicates it's 1 (there is no 1 in the dataset). The average here is 4.667, but if I calculate it elsewhere it's 6.67. I am not sure why this is happening. I've installed the latest version of JASP and this is the first thing I do.
Comments
This is really strange. I have tried to reproduce this but no luck so far. Is this a data set from the JASP data library? Also, this is clearly an issue for our GitHub page instead of the Forum (for details see https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/). Note that the mean from the descriptive table can not be correct either, if the frequency table is correct. But N=66 does match. Something unusual must be going on. If you provide the data (or any data libary example for which this goes awry) then I can check it out immediately.
EJ
Could the numeric stats use numeric values and frequency tables use labels? (E.g. value 1 has label 3). Just a guess, I'm on my phone
I've looked into this and I can reproduce this (e.g., in the JASP Data Library, the debug data set, the Descriptives table for the 0-1 contBinom variable falsely suggests the values are 1 or 2). This is a critical, hot-fix worthy bug. I have notified the team and will keep you posted. No idea how this passed our checks (or how it could even have arisen in the first place).
EJ
Hi Kevin,
Update: what happens is that JASP guesses the variable is "ordinal". For the "ordinal" and "nominal" scales, technically the mean is not meaningful. For these scales the different levels are assigned values starting from "1" and this explains the discrepancy. Setting the level of the variable to "scale" solves the problem (!).
This behavior will of course lead to confusion. In my opinion, JASP should present the mean etc. for ordinal variables (even though this is not strictly speaking meaningful); also, for nominal variables JASP should just report NA or NaN for the mean and other statistics. I am discussing this with the team.
E.J.
Hi EJ,
That did indeed solve the issue. I didn't catch that JASP changed that variable to an ordinal one. It's a little odd as the data for that variable are numerals only.
Kevin