search

cran-task-views / ctv

CRAN Task View Initiative
79 stars 13 forks source link

CRAN task view proposal: Epidemiology #25

Closed thibautjombart closed 2 years ago

thibautjombart commented 2 years ago

Motivation

The COVID-19 pandemic has made clear, once again, we need robust packages, built following best practices, Infectious Disease Epidemiology (IDE) (Jombart 2021). But the existence of good packages is just the first step in ensuring that they are used. The packages also need to be discoverable. Lack of discoverability leads to duplication of efforts and missed opportunities for collaboration. The existence of a task view summarising the main existing tools (in R) is an important first milestone to increase discoverability of R packages for IDE.

Scope

Packages included in this task view were identified through recommendations of IDE experts as well as an automated CRAN search using pkgsearch::pkg_search(), with the keywords: epidemiology, epidemic, epi, outbreak and transmission. The list was manually curated for the final selection to satisfy the conditions described in the previous paragraph.

Packages are deemed in scope if they provide tools, or data, explicitly targeted at reporting, analysing, modelling, or forecasting infectious diseases.

Packages

A current draft for the task view, with the proposed packages and their description is available at: https://github.com/Bisaloo/InfectiousEpi/blob/main/InfectiousEpi.md

Future updates will focus on adding more text and connections between packages to increase the added value of the task view relative to a simple list.

Overlap

We expect relatively little overlap with other task views excepted those focusing on general statistical and data analyses methods, which represent a large portion of the daily work of epidemiologists. Such general task views are for example Bayesian, ReproducibleResearch, Spatial or TimeSeries

Maintainers

The primary maintainer is Thibaut Jombart (@thibautjombart). Current co-maintainers are Hugo Gruson (@Bisaloo) and Matthieu Rolland (@mjrolland).

cjendres1 commented 2 years ago

In looking at the task views, I was surprised that there was no view for 'Public Health'. That would be a little more broad than 'Epidemiology', which is more broad than 'InfectiousEpi'. I would suggest the view be called 'Epidemiology', which is not much different in scope, more easily recognizable, and would aid in discoverability. In any case, there is a definite need for this task view.
I would like the maintainers to consider my package nhanesA for inclusion in this view. nhanesA provides ready access to the National Health and Nutrition Examination Survey (NHANES) data tables.

zeileis commented 2 years ago

Thanks Thibaut @thibautjombart for the proposal and apologies for responding so late.

Thanks also Christopher @cjendres1 for the initial feedback. I very much agree that "Epidemiology" would likely be a better scope. "Public Health" would probably become too broad, though.

Thibaut, maybe you can explain a bit better why you think that the scope should be limited to infectious disease epidemiology? I'm not an expert by any means in this field but my feeling is that the complement for general epidemiology wouldn't be so large. Also, I cannot see a good way how this complement could be covered well by another task view.

Similarly, regarding the inclusion criteria I'm also not sure when a general epidemiology package would or would not be deemed relevant for infectious disease epidemiology. Possibly, this impression is due to my lack of expertise but possibly sharpening the criteria would be possible.

Further comments:

mjrolland commented 2 years ago

Dear @cjendres1 & @zeileis,

I'll let @thibautjombart complete my answer, but the idea is that there are several fields of epidemiology that have relatively little overlap and require specific task views. For example I mainly work in the field of environmental epi, where I use hardly any of the packages/methods mentioned here, but many others that could (should...) constitute another task view. This is why we specified "infectious" epi, which was the scope of the TV.

I hope it makes sense.

Thanks for the feedback!

Matthieu

thibautjombart commented 2 years ago

Many thanks all for your input and feedback!

I agree there is a need for public health and non-communicable disease epidemiology to be covered in task views. I also fear 'Public Health' may be too broad, so focusing thoughts on whether the current proposal should be extended to general 'epidemiology':

My 2-cents is it would make sense to have a general 'epidemiology' TV if this can be done by adding a section or two to the current one; my fear is it might become very broad and loose some of the appeal of a TV. If we were to extend, it may be useful to have additional co-maintainers to better represent non-IDE work.

Let me know if you would like me to organize a call to discuss these matters live. But totally understandable if you'd rather keep the discussion here.

mjrolland commented 2 years ago

Thanks @thibautjombart,

My personnal opinion is that it would be more reasonable to keep this TV specific to infectious epi. The other fields of epi I know (environmental and genetic) have little to very little overlap with what is done in infectious epi, and I think would require their own task view rather than a chapter in this one.

However I am available to help if you decide to add an environmental epi chapter.

zeileis commented 2 years ago

Thanks for the discussion @thibautjombart and @mjrolland ! I like the idea of assessing how big the "delta" is between the current proposal and a more general "Epidemiology" task view.

Regarding the genetic side: In the transition to the new task view workflow we have retired the de-facto orphaned "Genetics" and "Phylogenetics" task views. There are currently efforts in establishing a new "Genetics & Genomics" task view. Possibly this could cover some of the genetic epidemiology stuff as well.

cjendres1 commented 2 years ago

It seems there is generalization within each tv, and it's fully expected that only a subset of constituent packages are relevant to any given application. Taking MedicalImaging as an example, there are packages for fMRI, EEG, PET, .. which are of no interest unless dealing with those specific images. There are also core image processing tools that are cross-disciplinary. That said, I believe it's best to keep the MedicalImaging tv as it is because there is value in being exposed to methods beyond your immediate interests. Just one example, but I feel that Epidemiology falls within a similar scope, and that a core group of packages can be identified with additional packages defined for specific applications. Perhaps you can begin with a comprehensive Epidemiology tv. If it appears that the scope is too broad, then at that point split into sub-disciplines.

thibautjombart commented 2 years ago

I like the idea. I would propose the following way forward:

Would this sound like a plan?

In addition I have 2 questions:

zeileis commented 2 years ago

I like this suggestion, it indeed sounds like a plan. Thanks!

Meanwhile I will try to encourage the other CRAN Task View Editors to also provide further comments/feedback.

Regarding the questions:

Thanks for your efforts!

cjendres1 commented 2 years ago

Sounds good to me! Thanks!

mjrolland commented 2 years ago

Hi all,

I have started a draft which I have shared with my lab mates, hopefully this should capture a good proportion of the main packages and sub-fields.

Also I'll add @privefl to this thread, he is currently writing the OMICS task view, which could overlap with environmental epi (eg methylation)

rociojoo commented 2 years ago

Hi everyone, Happy to see that everything is going on track, thanks for this proposal and the hard work! And thanks for anticipating possible overlaps with Omics, this should definitely be accounted for.

mjrolland commented 2 years ago

Thibaut is currently revewing a draft for the environmental epi section. It is clearly incomplete... I hope the TV users will take the time to add to it :)

davidjohannesmeyer commented 2 years ago

It seems there is generalization within each tv, and it's fully expected that only a subset of constituent packages are relevant to any given application. Taking MedicalImaging as an example, there are packages for fMRI, EEG, PET, .. which are of no interest unless dealing with those specific images. There are also core image processing tools that are cross-disciplinary. That said, I believe it's best to keep the MedicalImaging tv as it is because there is value in being exposed to methods beyond your immediate interests. Just one example, but I feel that Epidemiology falls within a similar scope, and that a core group of packages can be identified with additional packages defined for specific applications. Perhaps you can begin with a comprehensive Epidemiology tv. If it appears that the scope is too broad, then at that point split into sub-disciplines.

Fully agree. Personally, I would prefer more focused TVs which emphasize the most important packages instead of thriving for "completeness" not achievable anyway since TVs are moving targets by definition.

c3schmidt commented 2 years ago

A brief comment from an epidemiologist:

thibautjombart commented 2 years ago

@c3schmidt thanks for jumping in. This was my initial thought but based on feedback we collectively decided to work out a more general 'Epidemiology' TV with the possibility to split it eventually if it becomes too large.

thibautjombart commented 2 years ago

I have revised the proposed task view here.

This revision includes the following changes:

Many thanks to @mjrolland and @Bisaloo for their help.

mjrolland commented 2 years ago

Thanks! A colleague pointed out to me that in the "Spatial" TV there is a chapter on disease mapping: https://cran.r-project.org/web/views/Spatial.html#disease-mapping-and-areal-data-analysis, maybe there should be some reference to this?

rsbivand commented 2 years ago

Good idea. So far I'm unsure how one might link to a section in a TV.

zeileis commented 2 years ago

I could add a second argument section to view() so that view("Spatial", "Disease mapping and areal data analysis") generates the link that Matthieu used above. What do you think?

thibautjombart commented 2 years ago

I like the idea of being able to link sections in TVs. Meanwhile I have added the following paragraph to address @mjrolland 's comment: https://github.com/Bisaloo/Epidemiology/commit/429875fb32743d3a4e46b05583561b7b3d725834#diff-cad39e0cf72d6aba40137fa4a6d38ad296d90d6e5a6349b662cf2c4c48ecbecbR111-R117

For now linking to the TV and the section separately. Happy to update once the new feature is implemented.

@zeileis please let me know if there is anything else you'd like me to address for the proposal to be accepted

zeileis commented 2 years ago

So far I was just able to take a quick glance but in general it looks good. Some smaller issues I noticed include remaining BE ("visualisation" or "modelling") or two archived packages (oxcgrt, SCCS) where I'm not sure whether these should remain in the view for now or not. At least oxcgrt is unlikely to come back because it was archived for policy evaluations.

For the next step we would need endorsement of at least three CRAN Task View Editors. What do you think @rsbivand @eddelbuettel @rociojoo @davidjohannesmeyer @tuxette . A "thumbs up" reaction on this comment would be sufficient, I think, but more detailed feedback is of course also welcome.

rsbivand commented 2 years ago

OK, but note https://github.com/cran-task-views/ctv/issues/25, where some coordination might be worthwhile.

zeileis commented 2 years ago

Thanks, Roger, I agree. (But I guess you mean coordination with the data quality proposal, right?)

thibautjombart commented 2 years ago

So far I was just able to take a quick glance but in general it looks good. Some smaller issues I noticed include remaining BE ("visualisation" or "modelling") or two archived packages (oxcgrt, SCCS) where I'm not sure whether these should remain in the view for now or not. At least oxcgrt is unlikely to come back because it was archived for policy evaluations.

Many thanks for flagging these. Apologies, after 12y in the UK I am still confused about AE vs BE (but I just about managed to avoid French-English!).

I have just committed changes to use AE (visualization, modeling) and removed the two archived packages.

rociojoo commented 2 years ago

The TV looks great. My only minor suggestion would be to add examples in the Overview section Helpers subsection because miscellaneous tasks may be too ambiguous. They could be the same examples like those in the actual Helpers section: computing sample size, contingency tables, etc. You seem to be communicating well with the team behind the Omics TV proposal, which is great. I also agree with @rsbivand that there should be coordination with the data quality and official statistics teams. That's an endorsement from me, @zeileis

thibautjombart commented 2 years ago

@rociojoo thanks a lot for your input; now addressed in this commit

zeileis commented 2 years ago

Wonderful, thanks to my fellow editors for the endorsements and to Thibaut @thibautjombart (et al.) for doing all the updates so far. Further updates might be necessary to coordinate with task views under preparation (DataQuality and Omics). But I think we can already move forward with the current version. Next steps:

thibautjombart commented 2 years ago

Great! I have just closed the issue - had forgotten to close it, but all was indeed addressed. Will ping @Bisaloo to transfer the repos. Thanks all for your time, reviews and and input!

zeileis commented 2 years ago

Thank you for all your efforts, I look forward to seeing the task view on CRAN!

Bisaloo commented 2 years ago

Hi all, it looks like I cannot transfer / propose a transfer to an organization where I don't have any rights (source).

@zeileis, I invited you as a collaborator and will promote you to an admin once you accept and you should then be able to transfer the repo yourself.

Alternatively, I think you will need to create a team here and invite us first if you want me to do the transfer.

Please let me know what works best for you!

Bisaloo commented 2 years ago

Hah, I can't actually make anyone else admin for a repository hosted on a user account (vs an organization). I've transferred the repo to your personal account @zeileis.

zeileis commented 2 years ago

Thanks for your efforts! I know transferred it from my account to cran-task-views and everything seems to have worked smoothly. I'll streamline a few things to be consistent with our other task views. In particular, I will remove the proposal.md from the current revision and also drop the workflow because there will be the official CRAN workflow when everything is ready.

zeileis commented 2 years ago

OK, with my most recent changes I think we are good to go:

If I get green light from all of you (or simply a thumbs up for this comment), I will go ahead and make the CRAN release as well.

mjrolland commented 2 years ago

Awesome! Thank you :)

cjendres1 commented 2 years ago

Nice!

thibautjombart commented 2 years ago

Great! :)

zeileis commented 2 years ago

Fantastic! The task view is now live on CRAN (https://CRAN.R-project.org/view=Epidemiology) and also announced on Twitter (https://twitter.com/AchimZeileis/status/1534785592264904704).

zeileis commented 2 years ago

Just for the record: I have added the extended support for sections like view("Spatial", "Disease mapping and areal analysis") in the development version of ctv on R-Forge (to be released to CRAN in the next weeks). I've changed Epidemiology.md correspondingly.

thibautjombart commented 2 years ago

Awesome, thanks!