Error message 'list index out of range' and rectangular effect size graphs

[Heather Chen]

Hi there,

Thank you for making the web app for estimation stats. I came across the error message 'list index out of range' for the data set (Data 1) below when attempting to analyse with Hedge's g. I am aware that another person had the same problem in 2019 and the conclusion appears to be that it is due to a bug; could the error that I have encountered be caused by the same bug? With other samples, I have also encountered occasions where the effect size graph of an analysis performed with Hedge's g would be a rectangle (for example, for Data 2). I was wondering if you know what might have caused this?

Data 1

Control

0
0.72632944
28.5343709
33.7224384

Treatment

0
0
0
4.79896239
27.2373541

Data 2

Control

0.849090909

10

3.545454545

1.636363636

Treatment

0.180909091

0.262727273

0.203636364

1.727272727

4.090909091

Best regards,

Heather

[Joses Ho]

Hi Heather,

There are two separate issues here.

Firstly: where the full bootstrap curve is not displayed, you can mitigate this by adjusting the y-axis limits for the raw data (Step 6 for the two group plotting option on estimationstats.com).

Secondly, the `list index out of range` error is due to the data here being small (N<7) and more than half (3 out of 5) values are zero. There will thus be bootstraps where the resampled treatment will all be zeros. This will result in the pooled standard deviation being zero; as Hedges' g involves division of the mean difference by the pooled standard deviation, this leads to division by zero, and the observed error.

In lieu of increasing your sample size, you might consider rescaling or normalising your raw data, or reporting your effect sizes as mean or median differences.

Hope this helps,

Joses