Seventh Cross Validated Journal Club

Question

Time for the seventh iteration!

To remind for those who are new to the idea:

• CVJC is a meeting on chat where we discuss some paper and its theoretical/practical surroundings.
• It usually takes place on Friday about 18:00UTC.
• The paper must be OpenAccess or a (p)reprint suggested previously on a meta thread like this one and selected in voting.
• We will try to invite the author(s).

So, please suggest papers (each in one answer), and vote for your favourites!

EDIT: I see there are new suggestions coming, so maybe it is time to call the whole thing rebooted and set a new deadline to 23:59UTC 11.11.2012. For a fair voting, chl's suggestion gets -5 vote normalisation.

EDIT: Paper selected; as usual I'll try to invite the authors and then schedule the meeting.

EDIT: The 7th CVJC featuring Handling Missing Data by Maximum Likelihood by Paul D. Allison will take place on Friday 12.14.2012 at 16:00 UTC. You can register here.

Answer 1

I would like to propose the following article:

Paper 312-2012 
Handling Missing Data by Maximum Likelihood  
Paul D. Allison, Statistical Horizons, Haverford, PA, USA

http://www.statisticalhorizons.com/wp-content/uploads/MissingDataByML.pdf

Missing data seems to be an area in which all applied statisticians are, or should be, interested (I hope this is not a controversial statement!), and perhaps some theoreticians are interested also :-)

Answer 2

I would like to propose the following article:

Sheehan NA, Didelez V, Burton PR, Tobin MD (2008) Mendelian Randomisation and Causal Inference in Observational Epidemiology. PLoS Med 5(8): e177. doi:10.1371/journal.pmed.0050177

This paper is of some interest in relation to some threads on this site, in particular those related to inferring causality from observational studies.

Answer 3

I would like to suggest: A Better Lemon Squeezer: Maximum Likelihood Regression with Beta-Distributed Dependent Variables

This will be especially useful for DVs that have ceiling or floor effects or that are bounded, it is also useful when the variance may differ as a result of the IVs. But the paper gets kind of technical.