While reading the truly enlightening Data Analysis Using Regression and Multilevel/Hierarchical Models by Gelman/Hill and preparing for an introductory Bayesian course that is offered at FU Berlin this semester I decided to slowly switch to R more or less completely for all statistical day-to-day work.

There are some books and articles available specifically for beginners or switchers from SPSS, Stata or SAS, but the most valuable source is a collection of common use cases, including data management and graphs, freely available as Quick-R.

(via Dataninja)

A lot of useful tricks and tutorials on statistics and LaTeX are available on the Dataninja website. John also points out to this great Primer on LaTeX.

After reading through tons of working and conference papers lately, I have become slightly obsessed with the typographic insanities that are still widespread in social science where few people use proper tools for publication. These are the most annoying ones at the moment:

1. Double-spaced text. If you want your paper to be read, use line spacing that does not punish the interested reader. No, most conference papers, a lot of journals and even books are not typeset by a professional who could potentially use the huge space for annotations. Your paper is probably read like it is now, and anything above 1.4 spacing sucks most of the time.

2. Narrow margins, mostly combined with (1). If I was to scribble comments on your paper, I’d certainly use the right (or outer) margin for this. No margin, no comments.

3. Sans-serif fonts or Times. Unless you have found an exceptionally readable sans-serif, stay with serif fonts. Palatino, Minion Pro and others are freely (as in beer) available for LaTeX and word processors. Times is for two column layouts, which you should use only if you need to save paper!

4. “Figure 4 about here.” Am I supposed to cut and paste your tables and figures or constantly skip to the end of your paper? Unless you are preparing a manuscript directly for type-setting (and most of the time you don’t, see above), please leave your floats where they belong.

5. Endnotes. Footnotes are bad for references, but endnotes are just plain insulting. Don’t ever use them.

“But most journals only accept manuscripts like that!” Even if you often cannot discuss the usefulness of those requirements with the editors, most pre-publication articles (those that are circulated) can be made readable with a few clicks even in Word. It’s not that hard to have separate versions.

Yet another conference: The annual meeting of the communication research methods group in lovely Austria. The presentations were certainly better on average than at WAPOR. Our talk on online sorting studies was well-received, and there seems to be some interest in our super-secret sorting tool.

I especially enjoyed the talks on non-response in surveys (”The main difficulty in non-reponse research is non-response.”) and using SmartPLS for formative measurement models.

This year’s WAPOR conference was nice, although the papers presented varied (too) much in quality for my taste. The presentations by the commercial pollsters seem to be all about sample sizes, post-stratification weights and how to poll in Afghanistan, China or the world at large.

A quick review of the proceedings revealed that only about 8 papers used at least somewhat advanced methods of analysis like multilevel or structural equation models, latent class or time series analysis. There were some large-scale content analyses and cross-national comparisons done, but not many theoretical or conceptual advancements. Maybe WAPOR is not the place for cutting edge research, but then again what is?

Our own paper presentation went pretty smooth, we had nice co-panelists and I got to meet the head of Eurobarometer after complaining about their media use scales, oh well.