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

[Mixed Effect Models]: Simplify contrast specification #1583

Open cfrowd1 opened 2 years ago

cfrowd1 commented 2 years ago

Description

Making contrasts easier to specify

Purpose

More convenient--would no doubt help uaers who don't know about the sum averaging procedure.

Use-case

No response

Is your feature request related to a problem?

No

Describe the solution you would like

See above

Describe alternatives that you have considered

No response

Additional context

Perhaps also consider including different methods for specifying the Covaiance Structure.

juliuspfadt commented 1 year ago

Hi @cfrowd1, thanks for the request. Just so I do not get confused, your request is for a way to simplify specifying the contracts in a GLMM in JASP?

cfrowd1 commented 1 year ago

Yes, that's right, Julius, a simpler method of specifying contrasts. It can be difficult for some user to understand how these work (the patterns of +/- 1's and 0's) and a more straightforward way would be valuable. With best regards, Charlie.

-- Charlie Frowd CPsychol CSci AFBPsS Professor of Forensic Psychology, School of Psychology and Computer Science, University of Central Lancashire, Preston PR1 2HE.

EvoFIT >> www.evofit.co.uk

From: Julius Pfadt @.> Sent: 01 September 2022 9:20 AM To: jasp-stats/jasp-issues @.> Cc: Charlie Frowd <School of Psychology & Computer Science> @.>; Mention @.> Subject: Re: [jasp-stats/jasp-issues] [Feature Request]: Classical GLMM (Issue #1583)

Hi @cfrowd1https://checkpoint.url-protection.com/v1/url?o=https%3A//github.com/cfrowd1&g=MGQ5MjZmYmQ0MDlkYjQ3MA==&h=ZTAwZDcyMWQ1ZWI1YjMyYTU2N2YyMTJmODkwY2ExMjVlY2ZlNzI0MDUwZGY2MzlhOTFhN2QxZGY2ZDM4ZTkxZQ==&p=Y3AxZTp1Y2xhbmxpdmU6YzpvOjc0MjIxMTMwZTNlMWYxODIwOTk4MGNiMTlmNDNhMzc2OnYxOmg6VA==, thanks for the request. Just so I do not get confused, your request is for a way to simplify specifying the contracts in a GLMM in JASP?

— Reply to this email directly, view it on GitHubhttps://checkpoint.url-protection.com/v1/url?o=https%3A//github.com/jasp-stats/jasp-issues/issues/1583%23issuecomment-1233920674&g=NjBiYTgwMzM3OTZjMDk1NA==&h=Yjk4MDJiMDIzMDE3YzgzYzIwOGY4MjE0ZGU1NDkwMzMwNTQwNzQzYzVlNWEwMWE2ODRiNzc3OWZkY2NmYTVhZg==&p=Y3AxZTp1Y2xhbmxpdmU6YzpvOjc0MjIxMTMwZTNlMWYxODIwOTk4MGNiMTlmNDNhMzc2OnYxOmg6VA==, or unsubscribehttps://checkpoint.url-protection.com/v1/url?o=https%3A//github.com/notifications/unsubscribe-auth/AET4HGXXAOZXDODG3NDHNOLV4BRLFANCNFSM5MTLYRRQ&g=NDIwMDg5YTJlMjYzYjQ5Yw==&h=YjBlMWZlY2FjZGYxM2QxYjhmNWExYjYzYTEzYTk2NzlkZTNjOWRhMDNmNWZkMzgyNzlmZmQ0MjNlMGE5NzgwYQ==&p=Y3AxZTp1Y2xhbmxpdmU6YzpvOjc0MjIxMTMwZTNlMWYxODIwOTk4MGNiMTlmNDNhMzc2OnYxOmg6VA==. You are receiving this because you were mentioned.Message ID: @.***>

[University of Central Lancashire] https://www.uclan.ac.uk Please consider the environment before printing.

tomtomme commented 4 months ago

@cfrowd1 Do you have an idea on how to make it simpler? Any other software that does it better?

Maybe users could just click on the variables to contrast. They even could shift+click on multiple ones to group them. But if one would want to apply different weights for different vars, then this would still force the user to input a numerical value somewhere. And then we would be back, where we started from.

cfrowd1 commented 4 months ago

Good morning Thomas

I don't. All programs I've seen need an understanding of how to do it. I think it would be useful though to have a help button on the page to explain in plain English how to use them. That would be a nice improvement.

By the way, how can I find the estimates from the GLMM for intercepts and slopes for the random effects?

With best regards

Charlie.

-- Charlie Frowd CPsychol CSci AFBPsS Professor of Forensic Psychology, School of Psychology and Humanities, University of Central Lancashire, Preston PR1 2HE. www.uclan.ac.uk/academics/charlie-frowdhttps://www.uclan.ac.uk/academics/charlie-frowd EvoFIT >> www.evofit.co.ukhttp://www.evofit.co.uk/ EvoFIT on the 2022 Royal Institution Christmas Lectures: www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-bodyhttps://www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-body

From: Thomas Langkamp @.> Sent: 26 February 2024 11:04 AM To: jasp-stats/jasp-issues @.> Cc: Charlie Frowd (School of Psychology and Humanities) @.>; Mention @.> Subject: Re: [jasp-stats/jasp-issues] [Feature Request]: Classical GLMM (Issue #1583)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

@cfrowd1https://github.com/cfrowd1 Do you have an idea on how to make it simpler? Any other software that does it better?

Maybe users could just click on the variables to contrast. They even could shift+click on multiple ones to group them. But if one would want to apply different weights for different vars, then this would still force the user to input a numerical value somewhere. And then we would be back, where we started from.

— Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/1583#issuecomment-1963868545, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AET4HGUN6CQLZXCC7QAID5TYVRT2NAVCNFSM5MTLYRR2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJWGM4DMOBVGQ2Q. You are receiving this because you were mentioned.Message ID: @.***>

[University of Central Lancashire] https://www.uclan.ac.uk/ Please consider the environment before printing.

tomtomme commented 4 months ago

@cfrowd1 grafik

cfrowd1 commented 4 months ago

I thought that slopes could be included: there's a section to mention them, unless I've misunderstood. Any decent GLMM would want them to be there... With best regards, Charlie.

-- Charlie Frowd CPsychol CSci AFBPsS Professor of Forensic Psychology, School of Psychology and Humanities, University of Central Lancashire, Preston PR1 2HE. www.uclan.ac.uk/academics/charlie-frowdhttps://www.uclan.ac.uk/academics/charlie-frowd EvoFIT >> www.evofit.co.ukhttp://www.evofit.co.uk/ EvoFIT on the 2022 Royal Institution Christmas Lectures: www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-bodyhttps://www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-body

From: Thomas Langkamp @.> Sent: 26 February 2024 12:30 PM To: jasp-stats/jasp-issues @.> Cc: Charlie Frowd (School of Psychology and Humanities) @.>; Mention @.> Subject: Re: [jasp-stats/jasp-issues] [Feature Request]: Classical GLMM (Issue #1583)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

@cfrowd1https://github.com/cfrowd1 grafik.png (view on web)https://github.com/jasp-stats/jasp-issues/assets/4428185/4ea15bb7-89f9-4c97-b644-8b2c23837f43

This does not seem to include slopes however...

— Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/1583#issuecomment-1964037446, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AET4HGWKLFBYEE5EPYHMRT3YVR55VAVCNFSM5MTLYRR2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJWGQYDGNZUGQ3A. You are receiving this because you were mentioned.Message ID: @.***>

[University of Central Lancashire] https://www.uclan.ac.uk/ Please consider the environment before printing.

tomtomme commented 4 months ago

@cfrowd1 Its there when I do a gaussian model. The example from the data library above used a poisson model. grafik

cfrowd1 commented 4 months ago

Ah, that's in the latest model. Thank you!

Ok, for these contrasts then. It would be very helpful to have several options incl. at least simple contrasts.

For EMMEANS, could they calculate comparisons if an interaction is involved? At the moment, these are for only individual predictors.

With best regards, Charlie.

-- Charlie Frowd CPsychol CSci AFBPsS Professor of Forensic Psychology, School of Psychology and Humanities, University of Central Lancashire, Preston PR1 2HE. www.uclan.ac.uk/academics/charlie-frowdhttps://www.uclan.ac.uk/academics/charlie-frowd EvoFIT >> www.evofit.co.ukhttp://www.evofit.co.uk/ EvoFIT on the 2022 Royal Institution Christmas Lectures: www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-bodyhttps://www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-body

From: Thomas Langkamp @.> Sent: 26 February 2024 2:27 PM To: jasp-stats/jasp-issues @.> Cc: Charlie Frowd (School of Psychology and Humanities) @.>; Mention @.> Subject: Re: [jasp-stats/jasp-issues] [Feature Request]: Classical GLMM (Issue #1583)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

@cfrowd1https://github.com/cfrowd1 Its there when I do a gaussian model. The example from the data library above used a poisson model. grafik.png (view on web)https://github.com/jasp-stats/jasp-issues/assets/4428185/34027cb4-5226-4828-815c-d3e9b9f1ec05

— Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/1583#issuecomment-1964274928, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AET4HGRSZEJM4OCQI3UW7OLYVSLVJAVCNFSM5MTLYRR2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJWGQZDONBZGI4A. You are receiving this because you were mentioned.Message ID: @.***>

[University of Central Lancashire] https://www.uclan.ac.uk/ Please consider the environment before printing.

FBartos commented 2 months ago

Adding the "Help" button with a simple explanation in English is a good idea! I will check how to add that.

With respect to interactions, you specify contrasts on interactions using the +/- coding as you would do with main effects. The example in data library explains this using the Lark and Owls data set: image

cfrowd1 commented 2 months ago

Thank you. The problem with +/- coding is that it is complicated and people don't really understand it. Pushing a button that can do the contrasts for you would be ideal. This would make the software much more accessible. More of my students, for example, would be able to use it. Do the coding for the user... Just a thought. With best regards, Charlie.

-- Charlie Frowd CPsychol CSci AFBPsS Professor of Forensic Psychology, School of Psychology and Humanities, University of Central Lancashire, Preston PR1 2HE. www.uclan.ac.uk/academics/charlie-frowdhttps://www.uclan.ac.uk/academics/charlie-frowd EvoFIT >> www.evofit.co.ukhttp://www.evofit.co.uk/ EvoFIT on the 2022 Royal Institution Christmas Lectures: www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-bodyhttps://www.bbc.co.uk/iplayer/episode/m001gmxq/royal-institution-christmas-lectures-2022-dame-sue-black-3-living-body

From: František Bartoš @.> Sent: 11 April 2024 1:14 PM To: jasp-stats/jasp-issues @.> Cc: Charlie Frowd (School of Psychology and Humanities) @.>; Mention @.> Subject: Re: [jasp-stats/jasp-issues] [Feature Request]: Classical GLMM (Issue #1583)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

Adding the "Help" button with a simple explanation in English is a good idea! I will check how to add that.

With respect to interactions, you specify contrasts on interactions using the +/- coding as you would do with main effects. The example in data library explains this using the Lark and Owls data set: image.png (view on web)https://github.com/jasp-stats/jasp-issues/assets/38475991/be368480-3e00-4b8b-b630-07a5ee28513e

— Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/1583#issuecomment-2049562277, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AET4HGWVXZBQBQDYP2U5RETY4Z5BLAVCNFSM5MTLYRR2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TEMBUHE2TMMRSG43Q. You are receiving this because you were mentioned.Message ID: @.***>

[University of Central Lancashire] https://www.uclan.ac.uk/ Please consider the environment before printing.