Wednesday, August 25, 2010

LaTeX or not?

I am debating whether the everyday use of LaTeX for people writing documents that don't involve very dense mathematical formulation adds much value - most people accept that it is superior to packages like MSWord when you have a very high equation-to-word ratio. For some reason, people tend to get emotional about this issue when I raise it so keep the comments clean if you are going to comment. Beamer presentations made from LaTeX files are really great so lets not argue that one. But is the effort that a lot of people put into TeXing their documents worth it? 


Liam Delaney said...

some calmly discussed advantages and disadvantages below:

Liam Delaney said...

A more detailed but fervently pro-latex discussion below. hint - the title is "LaTeX: The way documents are meant to be written"

or how about "LaTeX: a genuinely better way to create scientific documents"

We were at a conference last year and one of the guys was presenting a fairly detailed paper. He had gotten the interest of the audience and a lot of questions. But the first question was simply a comment. After 25 minutes of hearing about the paper, the commentor put up his hand and when he got the speakers attention, he shoke his head, sighed and said in a reproachful tone "You used powerpoint..tuh"

Rob Gillanders said...

I was of the opinion that it was more hassle than it was worth to use Latex...but then I took the plunge and discovered the dirty little secret. Namely that it is really easy to learn - at least to a moderate level of sophistication. You're right that people get emotional about it. They see themselves as the keepers of a higher knowledge. I say use Latex but don't become an obsessive. Bibtex is fantastic too btw. A half a day with this Wikibook gave me nearly enough Latex-skillz to do my first couple of papers.

Liam Delaney said...

Would you advise an MA student to do their thesis in LaTeX Rob? My advice up to now has generally been no unless the person is naturally very techy.

Enda Hargaden said...

Deep breaths, deep breaths.

Word is fine for email-like correspondence, but so too is a hand-written note. For anything more substantive, LaTeX is preferable. The costs are higher but for many reasons (both obvious and subtle) the benefits outweigh the costs.

Word is to LaTeX as SPSS is to a good, clean Stata do-file.

The advantages to TeX are reasonably constant but the costs to creating a document can vary if you're not good with computers. (BTW these costs are lowered if you keep a template document that you can just bash away on, etc.)

So just as you might suggest to a second-year undergrad that SPSS is fine to calculate the means of his variables, anything over and above and he should probably learn Stata.

Liam Delaney said...

Ok Enda, I am going to do the SPSS/STATA thing at a later date. The comparison is not a great one. If I was trying to estimate a routine that wasn't available on SPSS then this would be a far bigger disadvantage.

If anyone has time or is bored enough maybe try to model the benefits of LaTeX. It might depend on the attributes of the user, the nature of the payoffs, the time costs involved in learning etc.,

Anyone who does do this model feel free to send it to me and I will post it on the blog. I will accept Word or LaTeX format!

Rob Gillanders said...

The MAs have a tight enough time constraint without adding Latex to their woes. Unless you get them messing around with it in the first semester. Plus the added value for one paper is probably not worth it so I'd only point them towards it if they were heading to Ph.D.-land

Liam Delaney said...

Or another one - if I were to randomly assign next year's masters class to a LaTeX condition what do you think the outcome would be for the treatment group? (assume away cross-contamination and all that for a moment).

Rob Gillanders said...

Well that would depend on whether their grade is dependent on how pretty their work is. I would argue that it shouldn't be at all (over the level of crayon scrawl) but I suppose Prof.s can have their heads turned by a sexy document.

Enda Hargaden said...

The random assignment for an MA thesis experiment overlooks the spreading of the fixed costs over many documents.

Liam Delaney said...

Fair point Enda - how about PhD initiation?

Liam Delaney said...

Rob - I'm leaving. For some reason your comments require me to click in (perhaps you don't have a gmail or something?). So it will be tomorrow if you are commenting again.

Colm Harmon said...

Thanks, Liam, for raising this. Some light-hearted thoughts, mainly written from the perspective of a supervisor of students. Nobody be offended!!

@Liam - would not recommend writing an MA thesis not least because the student should not have the time to learn LaTex amidst all of the other stuff they have to do....and outside of the academy nobody will care.

@Enda - totally OTT bordering on bull***t!! If you really want to take that idea forward than the analogy is with Gauss or R (or even old Fortran). Trust me - while you might think that Stata rules, most of the mega-wattage econometrics stuff that we read is actually programmed up from scratch in a primary programming language.

Now that I have my 'young man...' speech over with....(I love you really, Enda....)

I think most folks using LaTex in the wider academic (economics) world are using Scientific Word which is a halfway house of a package with lots of tweakability. I have it installed, I have used it, I can't see the value of it over Word in terms of ease of use right now particularly when things like citation managers etc are in such wide use. I use it when my co-authors use it. It takes no time at all to learn so no investment (as should be the case with a bloody word processor). As it happens most of them use Word so therefore so do I.

My ongoing problem with LaTex amongst the student cohort is sloppiness of the output, not the use of LaTex. That is not the fault of LaTex - it is the fault of the author. I need to see euro symbols, I need to see tables aligned properly, I need to see careful proofing, spell-checking. When a journal looks for a paper with 1" margins, double spaced throughout, yada yada yada I see folks getting all of a tizzy. I know that all of this can be done in LaTex - so go do it then. If you do not know how to do these things, then no matter how elegant the document looks it will amount to nothing. Use Word.

In terms of the equation writing, I would respectfully suggest that users of LaTex also need to learn the underlying notational conventions for using mathematics in documents before they start bashing them out in whatever package you are using. A matrix, for example, is written in BOLD ROMAN CAPS not normal italics. Equation Editor in Word is a good place to start if you want to learn these conventions :))

The final point - and this is a selfish one but hey, that's one of the perks of being a supervisor - is that you need to work in a manner that allows your supervisor to take your document and actually edit it. I am actually going to impose a rule on my graduate students that they use Word because (a) I can and (b) I want to be able to sit with their documents and work with them. And I am not going to learn LaTex!!

Colm Harmon said...

apologies - was getting weird error asking me to shorten comment - which I did - but clearly it was also just read the first one....the other four are the same only shorter!

Martin Ryan said...

@ Enda:

I agree that costs are lowered if one keeps a template document that will be used again. In addition to that, it is often also a case of "It's easy once you know how". That is, once you figure out how to do one particular thing, you wonder how you didn't know beforehand. In my experience at least.

I'll admit that the flip-side is full of disadvantage though. If one needs to make a very particular change to a document that one hasn't made before, it might not be possible to do it quickly. For me, this reason #1 as to why it would make sense to switch to Word.

@ Colm:

It might sound overly-technical, but the euro symbol problem occurs because the official euro symbol is meant to be a sans serif character, always the same regardless of the font being used. This violates normal typesetting design and practice: the dollar and pound signs are different for different fonts. So as you can guess, the euro currency symbol is not natively supported in LaTeX. One needs to install an add-on, and some of those available for download don't actually work.

(Anyone's who's interested should use "eurosym").

Moving swiftly on! Onto reason #2 as to why it makes sense to switch to Word: because a supervisor insists on it! One can't argue against this: if the switch to Word improves the ease of working on a shared document, then it's a no-brainer. One can also use Scientific Word if there is going to be a lot of equations; it's just a case of buying Scientific Word then.

@ Liam:

Given all of the above, and my appreciation of the valid reasons that there are for switching to Word, I think there is also a strong case for ditching Beamer - and switching to Powerpoint.

I think it's unfair to set Beamer off-limits when it uses much the same code as a LaTeX-generated paper. Creating a Beamer presentation is just as "involved" and requires just as much of a learning curve as creating a LaTeX-generated paper.

The main reasons for switching to Word are that:

(i) It can be difficult to learn new things quickly in LaTeX
(ii) LaTeX can be a barrier to collaboration between authors

These apply equally, if not more so, to Beamer. One can argue that it is even more important to change things quickly for a presentation. And to be able to let somebody else make those changes, if necessary. All of that could be tricky with Beamer.

The one thing that I have come to muse on about LaTeX-generated papers and presentations is that they need to be ready about a week before they are due. Otherwise there might be problems. So back to Microsoft then? Maybe Open Office? But that's another kettle of fish...

Enda Hargaden said...

@ Liam, yes I think random assignment at PhD level would yield benefits. The libertarian/adolescent rebel (aren't these the same thing?) in me thinks that it shouldn't be a rule, though.

@ Colm no offence taken! :) (Sure how could anyone take your views on technology seriously after 5x posting anyhow?) I agree with you on basic notation. Same of course applies to proper grammar, etc.

I think you overlook the problems with Word though. My friend had to submit her MA thesis on August 16. On August 15, she commented on Facebook "I'm not quite sure what just happened but I hate you, Word." The file wouldn't open. It's a very common problem - MS Word crashes far too often when documents are greater than about 30 pages. That can never happen with a .tex source file. Edits won't corrupt. (And btw you can edit .tex files in SWP.)

PS I started using R during my MA but Liam weakly advised against it! Point taken on the relative costs but imho the costs to learning LaTeX are much lower than the costs of learning R.

Peter Carney said...

...did anyone see my fur coat? it's getting nippy in here.

sorry liam.

My view: I dislike like LaTeX, in much the same way as I dislike shaving. But society has its demands, and so sometimes, just sometimes, I do both, admittedly poorly. I should also admit that most of Colm's anger is directed at me, for those of you who don't know Martin (and those that do but who struggle with deductive reasoning)

My new resolve: no more LaTeX for drafting working papers. Beamer for presentations. ...and a thorough examination of the determinants of shaving foam consumption amongst PhD students.. anyone with me? Dave??

Peter Carney said...

Aoccdrnig to rscheearch at an Elingsh uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are, olny taht the frist and lsat ltteres are at the rghit pcleas. The rset can be a toatl mses and you can sitll raed it wouthit a porbelm. Tihs is bcuseae we do not raed ervey lteter by ilstef, but the wrod as a wlohe.

Liam Delaney said...

A bit disappointing that few people have given the advantages of LaTeX so far. One real advantage is the importing of tables directly from STATA for example. So for example if you have a table programmed from a do-file when you compile the LaTeX document it will pull in the latest version of the table.

Also, there are obvious advantages if you are intending to self-publish something. For example, I would consider LaTeX for writing a set of lecture notes. But it seems

Martin Ryan said...


I think we can take an equal share of Colm's ire :-)


I don't think that the advantages of LaTeX (it looks better visually and is more effective at dealing with lots of equations) outweigh the disadvantages:

(i) It can be difficult to learn new things quickly in LaTeX
(ii) LaTeX can be a barrier to collaboration between authors

If everyone on a project is a LateX user, then I think there is case for recommending its use. Otherwise, I think problems will arise.

I could go on about:

(i) the magic border of 30 pages in Word

(ii) the awkwardness of Microsoft Equation Editor

(iii) the fact that LaTeX is widely used by mathematicians, scientists, engineers, philosophers, economists and other scholars in academia and the commercial world

(iv) That you can focus on the content of what you are writing
without being distracted by its visual presentation when using LaTeX

(v) That LaTeX Latex automatically numbers your sections, tables,
graphics, etc. It’s much easier once you get used to it

(vi) And finally, even though it sounds superficial, LaTeX looks a hell of a lot better, and that counts for something. The page “harmony” is better (due to the LaTeX page breaking algorithm); the interword spacing is better in
general; and page breaking is more intelligent. The “font expansion” feature means that the fonts are
being expanded or decreased to get better line breaking

However, despite all of the above, there is no getting away from the facts that:

(i) It can be difficult to learn new things quickly in LaTeX
(ii) LaTeX can be a barrier to collaboration between authors

In relation to the importing of tables directly from Stata, there is no advantage in using LaTeX, as far as I am aware. Getting content out of Stata is mostly done using outreg2 (for most folks, I reckon). There is an "outtex" command, but using outreg2 and specifying 'tex' instead of 'word' is actually more powerful. I have never heard of anyone pulling in data-output from Stata in the sophisticated way that you describe, but I know that it's possible with R (and Sweave) to have R files and LaTeX files hooked up in an automated way.

This actually brings me on to another point. When I use outreg2 to get regression-output from Stata into LaTeX, I still have a lot of work to do. To get the tables the way I (or others) want them to look requires a lot of messing around with LaTeX code, done manually :-(

if I was bringing output from Stata into Word, the formatting would be a lot quicker and easier. I definitely think now that it is too high a price to pay for superior style. If a Word doc is completely in Cambria font, and converted to Pdf, it looks good enough, IMO.

Martin Ryan said...

I will still use LaTeX for some things though. Papers and presentations are out due to the need to facilitate collaboration (and last-minute changes), but curriculum vitaes look much more impressive and are a lot easier to manage when made as a TeX file. I will still continue to do this.

Liam Delaney said...

if you've written the code for the table martin then surely updating the new numbers is less of a task than if you were starting an MS table from scratch?

Liam Delaney said...

ok forget what I said about emotional. Are there any LaTeX junkies out there who will make the case? Though real LaTeX junkies probably wouldn't be seen dead on blogger.

Martin Ryan said...

Simply putting in new numbers manually is one approach I hadn't considered Liam. However, this would be arduous - due to the dense nature of TeX code.

It's a good point though - that after my most recent bout of last-minute changes - I now have a table that I could potentially use over and over again, if I am prepared to wade through TeX code every time I run a different specification.

As it would be quicker to use Word though, I think it makes sense to plump for that.

Liam Delaney said...

Enda - I wouldn't advise people against R in a general sense but in the context of writing a three-month thesis where there has already been a lot of STATA training, my general advice is to stick with STATA. I hope someone does a post at some stage about the wonders of R.

Rob Gillanders said...

I make all my tables in Latex "by hand". (RATS doesn't seem to have an outreg). Takes very little time though of course more than getting outreg to do it would do. Still I like the sense of getting to know my numbers that it gives me.

Michael Breen said...

A view from a non-economist might be useful at this point :)

Migrating to latex is only worthwhile if you're using a lot of graphics and equations. It is far superior to ms word on both of these issues. Nevertheless, I keep seeing papers and beamer presentations (some in Geary!) where the author is not properly using latex's/stata's graphics capabilities.

On the issue of outreg2 and msword. I would echo Martin. I have also found it a lot easier to deal with stata output in word. It's less time-consuming period. Particularly if you are using outreg2 properly. Editing stata output takes only minutes in word. With hundreds of regressions in my thesis, it would have taken days to update the tables in latex if my supervisor wanted revisions.

So I find it puzzling that anyone who isn't using a lot of equations and graphics would bother with latex. I also find it puzzling that people with a lot of stata-output go to the bother of using latex. Finally, I find it annoying when people who go to the bother of writing their papers in latex don't properly use its capabilities... The amount of badly aligned tables, and messed-up graphs you see in latex-written papers these days....

Liam Delaney said...

Some good comments all the way through but I honestly don't think we have come near getting to the center of these issues. I will post again on this soon. For non-researchers this probably sounds trivial but people spend a lot of their time as researchers preparing and presenting documents so this issue is similar to any professional talking about their tools. Most of the really converted LaTeX users I know would deny any of the disadvantages above with the exception of the collaboration and fixed learning costs arguments (and they would usually dismiss these though this is a silly thing to do).

Eibhlin said...

@Martin With estout it is easy to create a table in Stata which automatically creates your LaTeX file and then you decide to add another control, drop an ob you just run your do file and the new table is created

Martin Ryan said...


I reckon there would still be manual labour involved in going through the TeX code to get the tables the way I want them to look. That is, I'd have to do a lot of tweaking in dense TeX code every time I run a new regression.

That would mean that I'd be much quicker using Word. And I think an important point emerging in this debate (as expressed by Michael) is that productivity is more important than style. But I will play around with estout and let you know. Thanks for the tip!


Enda Hargaden said...

@Martin - gotta agree with Eibhlin, I think you're being stubborn here. Estout/tabout to a separate file and tell LaTeX to include that file. Updates everytime you compile and it's about four lines of code.

On the point of substance over style: I agree. However I reach the opposite conclusion on TeX. Focusing on content is why I write up documents in plain text and leave the formatting to LaTeX. In word, I can't help but be distracted by the line-spacing/justification/font etc. It's far easier to ignore style when you're typing into a boring old ASCII notepad.

Martin Ryan said...

@ Enda: I actually never used estout before, or heard about it being particularly championed. But I will look into it when I get a chance!