Problem with dabestr median_difference
41 views
Skip to first unread message
Gianluca Finotti
Jun 2, 2021, 11:44:04 AM6/2/21
to estimationstats
Hi, I am getting very strange results each time I use the median_difference function (both in R and on the estimation stats website). More specifically, the confidence intervals around the median paired different appear to be wrong.
This is probably related to the fact that I get this error message when I use this function:
“extreme order statistics used as endpoints”. Am I right? If so, how can I solve this problem?
Please find a link to a google drive folder with an example of data and R code where I get this error: https://drive.google.com/drive/folders/1n3vFR01BObS05px6SvTAqiLEk4Phg9R1?usp=sharing
Many thanks in advance!
Gianluca
This is probably related to the fact that I get this error message when I use this function:
“extreme order statistics used as endpoints”. Am I right? If so, how can I solve this problem?
Please find a link to a google drive folder with an example of data and R code where I get this error: https://drive.google.com/drive/folders/1n3vFR01BObS05px6SvTAqiLEk4Phg9R1?usp=sharing
Many thanks in advance!
Gianluca
Adam Claridge-Chang
Jun 2, 2021, 12:00:53 PM6/2/21
to estimationstats
Hi Gianluca,
Thanks for the note. It appears there is a bug in the median difference function. There is another thread here about the same thing. We are investigating and hope to fix it.
Adam
Gianluca Finotti
Jun 2, 2021, 3:01:32 PM6/2/21
to estimationstats
Hello Adam,
thank you very much for your prompt reply. May I ask: is the same bug also present in the Matlab function?
Best,
Gianluca
thank you very much for your prompt reply. May I ask: is the same bug also present in the Matlab function?
Best,
Gianluca
Gianluca Finotti
Jun 3, 2021, 7:06:05 AM6/3/21
to estimationstats
Sorry again,
just one more question if you don't mind: I was using the median difference because I have to analyze non-normally distributed data.
Given that I don't know how long it will take to fix this bug, is there any other effect size or approach that you could recommend?
Thank you in advance.
Gianluca
just one more question if you don't mind: I was using the median difference because I have to analyze non-normally distributed data.
Given that I don't know how long it will take to fix this bug, is there any other effect size or approach that you could recommend?
Thank you in advance.
Gianluca
Gianluca Finotti
Jun 3, 2021, 3:53:51 PM6/3/21
to estimationstats
Sorry Adam, just one last question: do you have a rough idea of how long it might take to solve this bug?
I am asking because at the moment I have 2 manuscripts written and ready for submission that used the median_difference for all the analysis.
It is very important for me to have an estimation of how long it might take and whether I have to re-do all the analysis using a different approach.
Thank you in advance
Gianluca
I am asking because at the moment I have 2 manuscripts written and ready for submission that used the median_difference for all the analysis.
It is very important for me to have an estimation of how long it might take and whether I have to re-do all the analysis using a different approach.
Thank you in advance
Gianluca
On Wednesday, June 2, 2021 at 5:00:53 PM UTC+1 Adam Claridge-Chang wrote:
Joses Ho
Jun 4, 2021, 9:55:40 AM6/4/21
to estimationstats
Hi Gianluca,
The estimation stats web app and the Python package do not have any known discrepancies with paired comparisons; the bug is only in the R packages.
You can use either the web app or the Python package for your paired analysis for now.
So sorry for all the inconvenience caused!
Reply all
Reply to author
Forward
0 new messages