Updates based on surveymonkey feedback

[gndaskalova]

Hey Sandra, me again! @sandra-ab

I'm going through the feedback we got on surveymonkey over the summer so if you're happy to make some edits (and maybe you're editing anyways for DL stuff), I thought I can paste things here.

For the Intro to git and version control tutorial:

This tutorial didn't explain how to start with untracked files already made, because Rstudio puts all new projects in a home directory, which isn't the directory I have my untracked project which I want to now track. I also would appreciate explaining more details about the SSH component, because I made my repository private because my work is unpublished, and it complicated matters. Both my Rstudio and files are on a remote server I ssh into as well, which was confusing. I eventually found those answers elsewhere. But this tutorial was my main help!

[gndaskalova]

For the mixed effects tutorial:

When you first fit the mixed model you conclude that "once we account for mountain ranges, it's obvious that dragon body length doesn't actually explain the differences in the test scores", but I'm unclear how you came to that conclusion based on the summary output? It would be really great to see some kind of explanation of how to interpret that output.

Thanks for the great article on "Introduction to linear mixed models" I have just a silly question really. What does the 'er' in the lmer and glmer R packages stand for?

Potentially including examples with data extremes, like only categorical effects, or examples of non normal data

Lme4 is strongly inappropriate when it comes to analyse repeated observations, either short time or longitudinal. It doesn't allow us to choose the residual covariance, which is a must here. What it offers by default is either diagonal or compound symmetry, which are both plainly wrong unless well justified (eg just two observations per subject). This is banned for instance in drug research. To analyse repeated data one should use nlme, glmmTMB or glmmPQL and specify the proper residual covariance. It would be very worthwhile if you addressed this and informed the users to care of that. Too many people use wrongly lme4 for longer than 2-r repeated observations.

The code in the linked Github repository is not complete.

I'd love to see more examples worked, particularly for nested designs, as I have found coding these complicated. I'd also apprciate more information on how to lay out the results (the stargazer package part of the script did not work for me).

More discussion of the output of the models would be useful

[sandra-ab]

@gndaskalova Right thanks for those! Looks like the mixed effects model definitely needs updating!

I'll work on updating it with all the comments as a priority this week (the repeat data comment we received as an email as well). I think the stargazer package was updated (or scrapped?) a while ago which would explain the broken code.

The Intro to Git one I'm not so sure what to do - do you know how to answer these questions? I'm useless with Git and it would be nice to have it updated before the workshop next week!

[gndaskalova]

Hey @sandra-ab

Here is one for the Intro to Stan tutorial:

Two things: It is not clear which dataset you used, it would be nice if you could post a csv to your dataset. There is a bit of a hassle with the CXX vars installing stan for Windows, maybe post a link for windows users.

In case you can do these ones, the first point looks easily addressable at least.

For the git one, the SSH thing, I guess we could add a couple of sentences about what is an SSH link and a HTTPS link? The server part is quite user specific as not every person will be using a server. Maybe do a quick google on whether there is a helpful page online about using git with a server, then in our tutorial we can do just a "If you are using a server, check out this link".

Thanks!