1

I intend to deliver a short (1 hour) presentation to some interns and other staff (all of whom are quite statistically savvy, but are not statisticians -- mostly epidemiologists), on cutting edge statistical methods (especially if they are related to biostatistics, epidemiology/public health). I am hoping to provide a brief, non-technical overview of "cutting edge" methods. I am seeking any suggestions on topics that you think might be important to mention in this presentation. I'm looking for methods that may be quite familiar to statisticians, yet may not be to some graduate students with an epidemiology background for example. "Cutting edge" is also being loosely used here, since I intend to provide a brief overview of things like bootstrap and bootstrap aggregating techniques for variable selection and bootstrapping has been around for decades. It's just so important and might be familiar to these students, so I figured I'd include it. I also intend to present on propensity score analysis, a few machine learning techniques, etc.

So my question is, if you had to deliver a similar presentation, what other methods/techniques might you present given the audience and time constraint of 1 hour (my last slide might be a simple list of other important techniques we don't have time to cover, but that might be of interest to the students).

StatsStudent
  • 10,205
  • 4
  • 37
  • 68
  • 1
    How do you define "cutting edge"? What is the aim of your presentation? I guess that in 1h you could briefly mention maybe five methods without going into details, but I doubt you can do more... Moreover, this is **opinion based** since there is no definite list of "cutting edge" methods, so as such off-topic. – Tim Aug 09 '16 at 08:55
  • 1
    Not so much an answer, but as general advice it may be useful to look at recent issues of epidemiology journals for statistical methods papers and research articles which characterise their statistical methods as novel, cutting edge, innovative, etc. – Ian_Fin Aug 09 '16 at 09:15
  • 3
    http://stats.stackexchange.com/questions/6421/what-are-the-breakthroughs-in-statistics-of-the-past-15-years – caveman Aug 09 '16 at 09:35
  • Network meta-analysis is a technique which has moved from a research topic to the mainstream in health research recently. – mdewey Aug 09 '16 at 12:35

1 Answers1

2

Random causal forests paper by Athey and Wager is neat. ML for causal inference and heterogeneous treatment effects are probably of interest to epidemiologists.

Abstract:

Many scientific and engineering challenges---ranging from personalized medicine to customized marketing recommendations---require an understanding of treatment effect heterogeneity. In this paper, we develop a non-parametric causal forest for estimating heterogeneous treatment effects that extends Breiman's widely used random forest algorithm. Given a potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect, and have an asymptotically Gaussian and centered sampling distribution. We also discuss a practical method for constructing asymptotic confidence intervals for the true treatment effect that are centered at the causal forest estimates. Our theoretical results rely on a generic Gaussian theory for a large family of random forest algorithms, to our knowledge, this is the first set of results that allows any type of random forest, including classification and regression forests, to be used for provably valid statistical inference. In experiments, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially as the number of covariates increases.

dimitriy
  • 31,081
  • 5
  • 63
  • 138
  • I really, really like your idea of heterogeneous treatment effects. It's actually one of my areas of research too! Thanks for your suggestions, @Dimitriy V. Masterov. – StatsStudent Aug 09 '16 at 17:22