Problem with summary values if exponentials are used

[danyx23]

Hey there! I really like guesstimate and I use it to estimate some values related to the current coronavirus outbreak. I noticed however, that if you use a value from a distribution as an exponent that the summary values of that cell are not the correct values form the 50th, 5th and 95th percentile.

To reproduce: create a cell (10 to 20). Create another cell referencing this cell with (2^othercell). Check the summary values against the percentils in the expanded view. Below is a screenshot.

Let me know if I can help in fixing this. Thanks for a great app!

[danyx23]

I realized that what is shown in the big summary number is the mean, not the median. I think the median is the better number in this case - would you agree?

[OAGr]

Hi danyx23!

Glad to hear you like Guesstimate.

Correct that the number is the mean. I'm really not sure the median is better. In many cases people are interested in expected values and similar. I think the mean is generally more preferable, minus the fact that in cases with long tails (like the one above), the mean is pretty random, which is unfortunate.

Changing it without a big announcement would also be quite confusing to users who are used to it as is, so I'm hesitant to do so.

I recommend trying to organize things to not use long tails, Guesstimate doesn't handle them particularly well at this point.

[danyx23]

Thanks for your quick reply! I thought a bit more about this and I'm pretty sure the median is the more correct number to report - for gaussians etc it doesn't matter much but for skewed distributions the expected value is the one in the middle of all samples which is the median.

I can understand a hesitation to make this change for all old guesstimate documents, but maybe you could add a setting per guesstimate document that indicates which value is show? Or at least a preference if the scientific view should be enabled by default if people view this document?