ReplicationWiki:Village pump

From ReplicationWiki
Jump to: navigation, search

This page is for discussing operations of the site (features, policies, bugs, etc.).


Consolidating author pages with different spellings

I think that author pages such as Abhijit Banerjee and Abhijit V. Banerjee should be consolidated. Redirecting one to the other is easy, but how should one make a single page list all the papers with the multiple spellings? --Brian Quistorff (talk) 19:33, 13 June 2015 (CEST)

Best solution I can think of is redirect to the most frequent spelling and edit the code of the author template into the author page similar to what is done for the JEL code pages so that all spellings are accepted. The different spellings should also be listed on the main author page. J.Hoeffler (talk) 21:57, 13 June 2015 (CEST)
OK, I've modifed {{Article authors}} to allow up to two alternative names. You can see the results on Abhijit Banerjee ("Abhijit V. Banerjee" is now a redirect). --Brian Quistorff (talk) 01:09, 14 June 2015 (CEST)
That's great, thank you! J.Hoeffler (talk) 21:24, 14 June 2015 (CEST)
Do you also know how to list all versions of the name on the main page? J.Hoeffler (talk) 21:20, 16 June 2015 (CEST)
Done. --Brian Quistorff (talk) 19:56, 19 June 2015 (CEST)
Great, thanks! Only author for which we had more than three versions was Joshua Angrist, but not all four were actually used anyways, don't know how that happened. J.Hoeffler (talk) 23:36, 22 June 2015 (CEST)

Multilingual pages

For pages with multilingual content (e.g. Privacy policy, General disclaimer) do we want to split them up and then have a language selector somewhere on the page. Semantic MediaWiki has one at the bottom, for instance, but we could put ours at the top. Cheers. --Brian Quistorff (talk) 00:31, 14 June 2015 (CEST)

Not sure if they should be multilingual. The german and english version even somewhat contradict each other. I always had the plan to talk to our legal departmnet to check what we really need there but never got it done. J.Hoeffler (talk) 21:26, 14 June 2015 (CEST)

Shorter URLs

I think it would be a good idea to remove the "index.php/" from static (non-edit) URLs. Wikipedia and the like do this. Our URLs are already quite long with article titles. The can be done I think by the sysop (instructions at --Brian Quistorff (talk) 04:30, 15 June 2015 (CEST)

Requires server side modifications. Our administrator has been asked to update the wiki version in March so that we can make use of some new features... I am afraid we cannot expect much on that side before we get further funding. J.Hoeffler (talk) 18:08, 15 June 2015 (CEST)


PubPeer is a site for comments on articles. Would we want to systematically integration information from here on PubPeer? We could for instance post links to replications on the PubPeer page for the original article. Thoughts? --Brian Quistorff (talk) 04:46, 15 June 2015 (CEST)

Only journal I've seen there that we cover so far and that has one single short entry there is AEJ:EP (for a study that we have not yet integrated). There are also a handful of Econometrica and Review of Economic Studies papers that have comments. It seems to me it is not very much frequented by economists. Also, I don't like too much that one can comment anonymously there. There is so much still to do here, is it worth the effort copying content to another page? Maybe if it is just links to here we would get more attention. J.Hoeffler (talk) 18:18, 15 June 2015 (CEST)

XML dump of content

Is there a way to easily download all the data from this site for online analysis (and backup)? This would be nice given the aim is about transparency and openness:) Here is a guide for making an XML backup of the contents of a mediawiki site (I assume that'll work). There is a method if one has shell access and one that anyone can use. I'm not sure about the frequency we'd want, but Wikipedia appears to produce XML dumps about every month. --Brian Quistorff (talk) 21:05, 18 June 2015 (CEST)

There is a backup of the ReplicationWiki in the internet archive. After I found it I also added that information to the WikiApiary page. I don't really understand how that works, who did it and how it could be archived regularly. Once per month would be great in my opinion. J.Hoeffler (talk) 23:31, 22 June 2015 (CEST)
Yeah, I found that too. There are instructions for doing it using some code one can download. I'm running a dump right now. I'll post to the same page. --Brian Quistorff (talk) 04:50, 26 June 2015 (CEST)
Creating the dump is very easy. But, it turns out that another user can not upload a new dump to an page created by someone else. So, options include:
  1. Create a new page and upload dumps there. To prevent the same problem in the future, is there a "ReplicationWiki" email address that could be used as the account (so when I leave and others come someone else can upload to the same slot)?
  2. Post the dumps somewhere else. Is there non-wiki web space where these dumps could be hosted? I can host them on my website temporarily.
--Brian Quistorff (talk) 17:14, 28 June 2015 (CEST)
The email that gets the messages if someone applies for a new account is replication *at* Would it not be easier and make it less messy to ask the person who started the first archive? The account that was created for the email should also have web space but I would have to ask Jan Fahrenholz how to use it. J.Hoeffler (talk) 03:47, 2 July 2015 (CEST)

Comment fields

I wanted to enter, on a replication page, a link to the original author's reply to the replication. I think this would require a comments field in the form. I see sometimes this is mentioned in the raw page (not the form), so this encourages, or is this a temporary solution? I see that the comment field is mentioned on the Open tasks. Is there anything complicated about this addition (other things mention on that task were crossed out when added to the form but not the comments so I wasn't sure if this was harder). Are there limits to what people could write in there (e.g. links)? The display table is already pretty wide, is there a way that we can show the information on a narrower screen or in a vertical list? --Brian Quistorff (talk) 22:36, 18 June 2015 (CEST)

So far I did that underneath the form, and I tried to include the RePEc links because we get links back from them. Would be better to have an own field in the form for it. It is also not really defined what accepted/rejected means, this needs to be changed. Adding fields is not really hard. For smaller screens we think of a mobile version, and we already have an offer from a software firm to install it, just waiting for funding. J.Hoeffler (talk) 23:34, 22 June 2015 (CEST)

Using plural nouns in category names

Any objection to making category names consistently use plural noun forms (e.g. Category:Study lacking replication -> Category:Studies lacking replication)? --Brian Quistorff (talk) 22:40, 18 June 2015 (CEST)

I somehwat prefer singular but cannot really say why and JEL codes are also plural, so you are definitely right that we have an inconsistency. Why do you prefer plural? J.Hoeffler (talk) 21:59, 20 June 2015 (CEST)
Because the title describes the category contents, which is usually plural (the singular makes sense when looking at a page and seeing the category titles at the bottom). I think the standard is plural, that's what all the English Wikipedia (and associated projects). --Brian Quistorff (talk) 04:39, 21 June 2015 (CEST)
I see, plural seems to be standard in the English wikipedia version, singular in the German one, see for example Harry Rowohlt. An advantage of singular is that one could differentiate if the subjects are plural. J.Hoeffler (talk) 17:44, 21 June 2015 (CEST)
Interesting. Given we have an English interface, I'd say follow the English wiki convention for plurals, both for aesthetics (I think it's nicer for English speakers) and ease (other users are less likely to mess up). A multilingual interface for this wiki would be wonderful, and then it's really up to each language. The plural-of-plurals I don't think is a significant problem as it hasn't stopped the other English wikis, but we can always adapt if the need changes. --Brian Quistorff (talk) 22:46, 22 June 2015 (CEST)
It's not in the language that the decision was taken differently, it is a convention. Given the main categories (study lacking replication, replication, replicated study, journal...) are all singular so far this would require a lot of work and it's not really an important issue, so better save forces for other stuff, no? Up to now, no user has messed up anything with categories. Concerning multilingual interface, given so far the studies this wiki includes are all in English other versions only make somewhat sense to me as far as teaching material is concerned. There is some French an Spanish literature, and the Chinese publish studies in their own language (and we regularly have registrations from China, apparently because someone publishes about replication there and mentions the wiki and the presentations of my work in Nanjing and Toulouse) but I would say we can just wait until anyone wants to add anything about literature published in other languages. In German it is pretty much only literature on tax laws that are written in German that make sense to publish in German, other than that it is usually lower quality publications. This might be somewhat different in other social sciences for which the cultural context is more important like sociology or anthropology but again so far no one works on these here. J.Hoeffler (talk) 23:28, 22 June 2015 (CEST)

Root category

It might be nice to have all categories in the one unified tree so that there was a common root. I think this will help with finding categories and realizing consistent relationships. Common names are Contents or Fundamental. See Special:UncategorizedCategories. --Brian Quistorff (talk) 02:53, 19 June 2015 (CEST)

Made Category:Fundamental. --Brian Quistorff (talk) 05:09, 22 June 2015 (CEST)

JEL code pages

Some issues dealing with JEL categories:

  1. At the 1-digit level we have a few duplicates like N2 and Category:N2_-_Financial_Markets_and_Institutions. We want the later ones, right? Should we keep the former ones as just redirects or delete?
  2. At the 2-digit level there are both short titles (A19) and long (A11 - Role of Economics • Role of Economists • Market for Economists). We want the later, right? redirect/delete the other? It seems from dates that you were you in the process of moving from short name titles to long named titles.
  3. I think that the pages should just use {{Article JEL codes}} (rather than having all the wikicode on the page). I can modify the template so that it doesn't try to search for the full title as a property but just the first word (cf. User:Brian Quistorff/temp that splits the title apart by spaces and just uses the first term).

--Brian Quistorff (talk) 05:51, 19 June 2015 (CEST)

N2 should redirect to Category:N2_-_Financial_Markets_and_Institutions.
At the 2-digit level the move from simple codes to code with current definition has not yet been completed. If you can modify the template this would make the move much less tedious. What complicates the thing is that some codes were used that do not exist. It might be that they existed in the past, I have not checked how the codes changed over time, it is not documented on the website or in the internet archive under the current page. Especially letters with a single 0 were sometimes used although for those letters only a 0 with a second digit is currently defined. J.Hoeffler (talk) 15:37, 21 June 2015 (CEST)
Should we redirect codes like D0 to D00? The D00 page could list both. --Brian Quistorff (talk) 05:56, 22 June 2015 (CEST)
For letters like B, H, and I that only have 00 this would be good, but D0 could refer to anything from D00 to D04, so redirecting to D00 is not correct. How about we list them under D and mention that they were under D0 although that does not exist? Would that also be a solution for the other code that does not (or no longer) exist, D27, list it under D2? J.Hoeffler (talk) 23:16, 22 June 2015 (CEST)
Yeah, that sounds good. --Brian Quistorff (talk) 04:41, 25 June 2015 (CEST)
Done. J.Hoeffler (talk) 17:25, 4 July 2015 (CEST)

Bot usage

As there are many JEL pages to clean up, I was thinking of using a bot. What do people think the process for this should be? From looking at some other wiki projects the basic process is to create a separate user account, run a few test edits, get approval from the community (e.g. discussing it here), and then getting a "bot flag" from a bureaucrat (so that the edits don't appear in Recent Changes). Operation guidelines include:

  • Don't make changes too quickly as this may slow the server.
  • Stop if received a message from admins or others.
  • Make uncontroversial edits.
  • Fix your own mistakes.

Do these seem good? Anything else? Likely there will be a need for similar tasks in the future. --Brian Quistorff (talk) 22:40, 22 June 2015 (CEST)

There already is a bot account, ReplicationBot but the guy who ran it is no longer really working for the project other than small and important tasks. If we can get a bot to clean these pages up it would save a lot of time. Your suggestions seem very good to me. J.Hoeffler (talk) 23:20, 22 June 2015 (CEST)

Making nice Author pages

The author pages are not really easy to look at as each bullet item is just a mash of fields and values. We can, however, pass the data from each entry through a template to format it intelligently. I think we should do that. I've mocked up a simple version of one search result list here (compare to top list at the current page). Feedback on the layout is welcome. --Brian Quistorff (talk) 04:58, 26 June 2015 (CEST)

Your version is a great improvement, we should use that format for all the pages that list studies! Authors is probably so far the least important one given that since they are not yet even nearly complete they are not yet linked from the main page. Should we change that? I still hope for Christian Zimmermann to give us the backlinks from RePEc for these, then we will get much more attention. J.Hoeffler (talk) 10:55, 26 June 2015 (CEST)
I started first with the author pages because they seemed the simplest. I'll put this change into the author template.
I agree that we should do something for the study pages as well. The top two sections (bibliographic info and study materials/methods info) should be easy to change and I'll start a mockup. The table of replications seemed troublesome to me. It's really wide, so hard to use. I could redesign it as a list (like the authors page) but then one couldn't sort columns. Is that a feature that we really want to keep? --Brian Quistorff (talk) 04:15, 27 June 2015 (CEST)
I meant your template version could be used for the JEL codes, methods and software templates, too. J.Hoeffler (talk) 16:17, 30 June 2015 (CEST)
Ah, yes. I'll put it on the to do list. --Brian Quistorff (talk) 21:57, 30 June 2015 (CEST)
OK, I've used the same list formatting for those other pages (JEL codes, keywords, methods, data types, etc.). --Brian Quistorff (talk) 16:38, 4 July 2015 (CEST)
That is really great, thank you! J.Hoeffler (talk) 16:00, 7 September 2015 (CEST)

Custom Javascript

I was wondering if there is any way to load custom Javascript? I wanted to load a script to convert timestamps to local time in discussions. I tried using my user page (User:Brian_Quistorff/common.js) but since that didn't work I think that $wgAllowUserJs must be set to false on this wiki. It also looks like Gadgets are not enabled (not listed in Special:Version). --Brian Quistorff (talk) 04:57, 28 June 2015 (CEST)

I am trying to get new funding to have a specialist work on such issues, at the moment unfortunately we have no one who is both able and willing to do this. J.Hoeffler (talk) 16:02, 7 September 2015 (CEST)

Traffic lights

Are traffic lights still set by bots? Shouldn't they just be set by the templates (which should be possible)? Is there a reason many normal pages in Category:Replication and Category:Replicated study don't have them set (did the bot stop?)? --Brian Quistorff (talk) 02:22, 29 June 2015 (CEST)

Some don't have it if there is more than one replication and it is unclear which light should be set. Not sure if there are still bots at work, would have to ask Jan Fahrenholz. J.Hoeffler (talk) 03:42, 2 July 2015 (CEST)

Bot fixing of JEL code pages

As a check, I believe we eventually want all the JEL pages as category pages with page names that include the JEL titles. So here is my proposed plan for fixing them up:

  1. Bot generating the missing categories (or ones that exist but are redirects). Have a look at my test edits. If they look fine, as a bureaucrat can you give my bot the "bot status flag"? (It isn't really that necessary as few people are watching Special:RecentChanges, but it seems like the right procedure.) Then, I will continue making them, starting with the 3 character codes and then doing the 2 character codes. Then we should have all the pages as category pages.
  2. Bot delete the JEL code pages that are either not in the Category namespace or are the shorter version (like "L00" rather than "L00 - title").
  3. AWB cleanup the pages that already had been made (like removing the now unneeded {{attention}} tags talking about "empty not being empty").

How does that seem? --Brian Quistorff (talk) 20:28, 29 June 2015 (CEST)

The bot account is already set to be a bot account, what else needs to be done? The JEL codes pages that just consist of the code are necessary as redirects so that in the future users can click on every page on the code and get to the respective category. As for the template tags, it's good to remove them as it no longer says "empty" but the brackets are just empty and in the long run it would be better to indicate that this says nothing about how much they contain. Would be great to have a number there like on the main page for number of replications in the wiki. J.Hoeffler (talk) 16:15, 30 June 2015 (CEST)
OK, I'll leave the redirects.
I was also think of making a set of templates to programmatically go from code -> title. For instance, we could have {{L10-title}} be the title text for that JEL code. That way if we pass a code into a template we could get to the category by using [[Category:{{{1}}} - {{{{{1}}}-title}}]]. Do you have a preference on what format to name these templates: L10, L10-title, or something else? --Brian Quistorff (talk) 22:05, 30 June 2015 (CEST)
It's fabulous you completed the category tree! Now we should make the codes clickable on all pages! I am not sure if I understand what you want to do with those templates. You want all study pages to be in the categories of their code? That would solve the problem with the empty code categories. I would say those categories should be hidden on the study pages in order to avoid redundant information. On the study pages I strongly prefer short codes, otherwise they take too much space. J.Hoeffler (talk) 03:17, 2 July 2015 (CEST)
My other ideas was a bit different, but I'll hold off either way until there's a need for it. I like your idea about linking the JEL codes. We should do that. (I think again using templates to format the outut). --Brian Quistorff (talk) 10:58, 2 July 2015 (CEST)
If we make the JEL codes automatically clickable they link to the page with the same name, which currently redirects to the JEL code tree. I have to check how I did that once, it has to do with how the variables are called in the form if I remember correctly, got the inspiration for it when I asked a question in a semantic media wiki forum and was shown like in this other wiki as an example. The changes I did at the time can still be seen in the history, seems to me it was mainly to Template:Article/replication/entry. At the time it didn't make too much sense here because most of the JEL code pages still did not exist. I guess for authors this would be very nice, too, and it would get us a lot of attention if we get backlinks from IDEAS/RePEc for it. Then doing the same with software and methods seems helpful to me but their fields need to be reformatted because they still often have content that does not make sense to be clickable (software versions, descriptions of methods). Geographical origin of data would also be nice to have clickable. Probably data type would be easiest because that one is standardized. Oh, and keywords, obviously. J.Hoeffler (talk) 14:00, 4 July 2015 (CEST)
Given that the lists you created are much better than just titles of the articles if we get all of them into the category tree we have the problem that we do not want to show them there but rather keep the lists, would that be possible? J.Hoeffler (talk) 21:48, 6 July 2015 (CEST)

Concordance between study and replication pages

For an study-replication pair is some process by which the study page and the replication page show the same information? For example, if a user enters information on a 'replicated study' page, is there some process to ensure that the 'replication' page is created? And the information matches (even if one is updated)? --Brian Quistorff (talk) 21:25, 29 June 2015 (CEST)

No, but it would be great to have that. We already have two replications for which the pages of the original studies were not created. And there may be inconsistencies. J.Hoeffler (talk) 03:33, 2 July 2015 (CEST)
If each replication page lists the titles that it replicates (via an arraymap like we do for JEL codes), why doesn't each study page just list all replications by using {{#ask:[[replicated_title::{{title}}]] rather than having the information re-entered? Replication pages could also list the studies it replicates by knowing just the titles, rather than having all the info re-entered. This would help (a) eliminate inconsistencies and reduce effort to enter material, (b) allow studies with multiple replications and replications of multiple studies to be edited by forms, and (c) easily allow converting a 'study lacking replication' to a 'replicated study' (just enter the replication and it would automatically show up on the study page). As we get more pages in the wiki, (c) will become more likely. We could make a parallel set of forms and templates to try out on a few pages to see if there are any problems with this approach. --Brian Quistorff (talk) 16:55, 2 July 2015 (CEST)
Point b) would be a very significant improvement because these are the most interesting pages. Arraymaps so far work with comma separation, that is not feasible for titles because they sometimes contain commas themselves. And so far they are not even properties. Multiple instance templates would be much better. For those we need a software update and an extension. J.Hoeffler (talk) 14:03, 4 July 2015 (CEST)
A few possible solutions:
  1. We could use a different arraymap delimiter that wouldn't be in titles (maybe @ # or ~).
  2. We could allow linking a replication to studies via another property than the title (doi, repec id, etc.)
  3. Since we only need one piece of info to link to original studies (e.g. title) then we could probably handle most cases feasibly by allow a form with say 10 fields (title1, title2, ..., title10). The replications I can think of that have more than 10 are are meta-analyses that also replicate the original studies to provide more uniform estimates (example) but my guess they are few.
--Brian Quistorff (talk) 18:50, 5 July 2015 (CEST)
The forms are nice and easy but already separating everything with commas and capitalizing is too much for some users. Separating various titles with another sign is just not nice. We need those multiple instance templates, that is the much cleaner solution. I also find the title the most intuitive property to use as a link to a study. DOI and RePEc ID should of course link to the DOI or RePEc pages. I wholeheartedly agree that it would be extremely helpful to combine entering replication and replicated study information in one step to avoid redundancy and differences in information that should be identical. J.Hoeffler (talk) 21:50, 5 July 2015 (CEST)


When I went through the open tasks and updated them next to the clickable properties I found the improved search a very important feature because the ask form is horribly user unfriendly and so far the only real way to do combined searches that most users will be interested in - like "which studies have data and code in the software I use in my teaching and the method I want to show". We created our own form but it does not really work, I think it only finds replications or replicated studies because some of the fields only exist in these. Maybe we could fix that somehow? J.Hoeffler (talk) 14:14, 4 July 2015 (CEST)

Making nicer Article pages

I've taken a stab at making article pages look nicer. Compare an existing article to a reformatted version. It's about 2x more compact (usually all info is visible initially). It also looks fine on narrower screens as it doesn't have the old tables that couldn't get any narrower. What do you think? --Brian Quistorff (talk) 12:50, 5 July 2015‎ (CEST)

To be honest I prefer the old format: With the tables everything is well defined right at the place like at every other article and like in the description that one can access clicking on the headings and even going with the cursor over them. And your changes in the templates have destroyed the journal issue categories because the brackets were lost (fixed it). Furthermore, I think we should prioritize better what to work on. I see your point with the smaller screens but for that I would say a mobile version is more important and so far I would guess not so many users would access the wiki with a small screen anyway. Our focus group is scientists who work with software like R or Stata, you don't think of them working on a smartphone. Hm, maybe for a mobile version the tables wouldn't work well anyways? Probably it would be possible to keep the clickable headings and the description one gets when going with the mouse over them also without table? Definitely the data/code/readme information should stay together because one is usually jointly interested in them. J.Hoeffler (talk) 21:43, 5 July 2015 (CEST)
Transformed it a bit, made some comments and checked on my phone. Indeed has the advantage one does not have to enlarge everything to make it readable and more of it fits on the screen. I guess just for checking a study people might indeed once in a while use mobile devices to access the wiki. And should be possible to get the hoverboxes also in the new format. J.Hoeffler (talk) 17:13, 6 July 2015 (CEST)

Unused properties

Does anyone know what to do with these? I don't get what the list means, it seems to indicate there are errors at certain pages but when looking at those pages they seem completely normal to me. J.Hoeffler (talk) 16:05, 7 September 2015 (CEST)

Personal tools