search

jasp-stats / jasp-issues

This repository is solely meant for reporting of bugs, feature requests and other issues in JASP.
55 stars 29 forks source link

[Bug / Feature]: Update Highlights for Negative Binomial #1663

Open levolz opened 2 years ago

levolz commented 2 years ago

JASP Version

0.16.1

Commit ID

No response

JASP Module

jaspDistributions

What analysis are you seeing the problem on?

Negative Binomial

What OS are you seeing the problem on?

Windows 11 (& others)

Bug Description

Highlighting the CDF of an interval includes the lower bound, i.e. difference between both cdf highlights does not equal the cumulative probability over the interval.

Expected Behaviour

The lower CDF highlight should give the [0, a -1] cdf instead of [0, a] cdf.

Steps to Reproduce

Highlighting CDF in the negative binomial analysis in the cumulative distribution plot

Log (if any)

No response

Final Checklist

levolz commented 2 years ago

Thanks for the explanation! The PDF values within the discrete probability plots are indeed a bit confusing :D

However, I mean that the CDF highlight in the PDF plot should be the difference between the CDF highlights in the CDF plot (i.e., the issue concerns the CDF plot). Hope the picture below illustrates what I mean better than the words - i.e., if below you'd show the CDF(a - 1) = 0.19, they'd add up, while currently, the CDF(a) does not really add informational value to the analysis.

[image: grafik.png]

On Thu, 17 Mar 2022 at 17:28, Simon Kucharsky @.***> wrote:

Thanks for filing the report!

difference between both cdf highlights does not equal the cumulative probability over the interval.

I think the problem is that there are no two cdf highlights. There are two probability masses shown, and one cumulative probability shown. The two values at the end of the "highlights" merely show the probability mass of the value of the random variable, not the cdf. The value associated with the "cumulative probability" concerns the whole interval that is colored, and specified according to the interval in the options (which indeed includes both the lower and upper bounds) [image: Screenshot 2022-03-17 at 17 19 23] https://user-images.githubusercontent.com/7093546/158845950-70c81f18-88db-4743-ab7b-ca5f90970471.png

In other words, the value of the cumulative probability is the sum of the probability masses of all values within the specified interval, not only the lower and upper bounds.

Does that clarify the issue?

I see that the options/plots are somewhat confusing for the discrete distributions. The concept of the plot borrows from the plots in the continuous distributions, where the distinction between density at a specified point vs. the probability mass inside of the interval is easier to spot, whereas for the discrete distributions it is perhaps confusing to understand what part of the plot relates to the cdf and what to the pmf. I would be happy to hear suggestions on how to clarify/change this.

— Reply to this email directly, view it on GitHub https://github.com/jasp-stats/jasp-issues/issues/1663#issuecomment-1071051893, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKDMDDTVYYSAHGWXXEIKJGTVANMUBANCNFSM5Q7FOB3Q . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you authored the thread.Message ID: @.***>

levolz commented 2 years ago

CDF(a-1) = 0.11, ofc. - my mistake

On Thu, 17 Mar 2022, 17:50 Leonhard Volz, @.***> wrote:

Thanks for the explanation! The PDF values within the discrete probability plots are indeed a bit confusing :D

However, I mean that the CDF highlight in the PDF plot should be the difference between the CDF highlights in the CDF plot (i.e., the issue concerns the CDF plot). Hope the picture below illustrates what I mean better than the words - i.e., if below you'd show the CDF(a - 1) = 0.19, they'd add up, while currently, the CDF(a) does not really add informational value to the analysis.

[image: grafik.png]

On Thu, 17 Mar 2022 at 17:28, Simon Kucharsky @.***> wrote:

Thanks for filing the report!

difference between both cdf highlights does not equal the cumulative probability over the interval.

I think the problem is that there are no two cdf highlights. There are two probability masses shown, and one cumulative probability shown. The two values at the end of the "highlights" merely show the probability mass of the value of the random variable, not the cdf. The value associated with the "cumulative probability" concerns the whole interval that is colored, and specified according to the interval in the options (which indeed includes both the lower and upper bounds) [image: Screenshot 2022-03-17 at 17 19 23] https://user-images.githubusercontent.com/7093546/158845950-70c81f18-88db-4743-ab7b-ca5f90970471.png

In other words, the value of the cumulative probability is the sum of the probability masses of all values within the specified interval, not only the lower and upper bounds.

Does that clarify the issue?

I see that the options/plots are somewhat confusing for the discrete distributions. The concept of the plot borrows from the plots in the continuous distributions, where the distinction between density at a specified point vs. the probability mass inside of the interval is easier to spot, whereas for the discrete distributions it is perhaps confusing to understand what part of the plot relates to the cdf and what to the pmf. I would be happy to hear suggestions on how to clarify/change this.

— Reply to this email directly, view it on GitHub https://github.com/jasp-stats/jasp-issues/issues/1663#issuecomment-1071051893, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKDMDDTVYYSAHGWXXEIKJGTVANMUBANCNFSM5Q7FOB3Q . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you authored the thread.Message ID: @.***>

Kucharssim commented 2 years ago

Not sure why my previous reply is deleted...

(i.e., the issue concerns the CDF plot).

You are right. I was only looking at the PDF plot.

tomtomme commented 5 months ago

@levolz As far as I can see in JASP 0.18.3 everything seems fine with the cumulative density function - if this is the CDF you were even talking about. Can you confirm and then close this issue?

levolz commented 5 months ago

No, I think the issue persists (if I am not mistaken) - see screenshot: image

The issue was referring to CDF(2 <= x <= 5) = CDF(x <= 5) - CDF(x < 2) ~= .98 - .75 = .23 --- however, the highlighted CDF values are .98 (upper) and .88 (lower), which conceptually does not make sense - it should instead be at the lower end of the highlighted PMF(x = 2) = .13, corresponding to CDF(x < 2) = .75

tomtomme commented 1 month ago

still valid for latest 0.19beta