Already some weeks ago I have finished up some thoughts for a Replication Bayes factor for ANOVA contexts, which resulted in a manuscript that is available as pre-print at arXiv. The theoretical foundation was laid out before by Verhagen & Wagenmakers (2014) and my manuscript is mainly an extension of their approach. We have another paper coming up where we will use it to evaluate the success of an attempted replication of an interaction effect.
Pre-registrations are becoming increasingly important for studies in psychological research. This is a much needed change since part of the “replication crisis” has to do with too much flexibility in data analysis and interpretation (p-hacking, HARK’ing and the like). Pre-registering a study with planned sample size and planned analyses allows other researchers to understand what the initial thinking of the authors was, how the data fits to the initial hypothesis and where are differences between the study protocol and study results. In theory, it looks very simple: you think about a problem, you conceive a study, lay out the plan, register it, collect data, analyse and publish.
This is an interesting article from The Guardian on “post-truth” politics, where statistics and “experts” are frowned upon by some groups. William Davies shows how statistics in the political debate have evolved from the 17th century until today, where statistics are not regarded as an objective approach to reality anymore but as an arrogant and elitist tool to dismiss individual experiences. What comes next, however, is not the rule of emotions and subjective experience, but privatised data and data analytics that are only available to few anonymous analysts in private corporations. This allows populist politicians to buy valuable insight without any accountability, exactly what Trump and Cambridge Analytica did. The article makes a point how this is troublesome for liberal, representative democracies.
Nathan Silver’s FiveThirtyEight has had an excellent coverage of the US Presidential Elections with some great analytical pieces and very interesting insights in their models. Each and every poll predicted Hillary Clinton to win the election and FiveThirtyEight was no exception to that. Consequently, there was a lot of discussion on pollsters, their methods and how they – again after “Brexit” – failed to predict the outcome of the election. There are many parallels between the elections in the US and the Brexit-vote in the UK. At least for the US, however, the predictions weren’t that far off. And FiveThirtyEight in particular, gave Trump better chances than anyone else:
For most of the presidential campaign, FiveThirtyEight’s forecast gave Trump much better odds than other polling-based models. Our final forecast, issued early Tuesday evening, had Trump with a 29 percent chance of winning the Electoral College. By comparison, other models tracked by The New York Times put Trump’s odds at: 15 percent, 8 percent, 2 percent and less than 1 percent. And betting markets put Trump’s chances at just 18 percent at midnight on Tuesday, when Dixville Notch, New Hampshire, cast its votes.
Just a short post on a new paper that is available from our department. If you happen to have calculated factor score estimators after factor analysis, e.g. using Thurstone’s Regression Estimators, you might be interested in the reliability of the resulting scores. Our paper explains how to do this, compares the reliability of three different factor score estimators and provides R- and SPSS-scripts for easy estimation of the reliability. While some reviewers have argued, that this reliability cannot exist, I think, we have some good arguments how our perspective is in line with existing literature on psychometrics.
The paper is available as Open Access in the International Journal of Statistics and Probability and you can find the article here. I have uploaded the scripts to GitHub, so you can easily download them, add issues or create forks. The repository is at https://github.com/neurotroph/reliability-factor-score-estimators.
In recent years, it has become a notion to not only report point estimates of effect sizes, but also confidence intervals for said effect sizes. I have created a small R script to calculate the bounds of such a confidence interval in the case of t- and F-distributions.
The New York Times published an interesting piece on the differences between pollsters’ predictions. All five predictions used the same data set, so sampling differences are not of concern. Still, there was a difference of up to 5% between the predictions.
Michael Inzlicht has posted an article on his blog about how he lost faith in psychological science after reading the now infamous paper on “false-positive psychology” 1.
Already in September last year, Der Spiegel published an interview with Peter Wilmshurst, a British medical doctor and whistleblower who made fraudulent practices in medical research public:
In the course of the 66-year-old’s career, he conducted studies for pharmaceutical and medical devices companies, and unlike many of his colleagues, never hesitated to publish negative results. He’s been the subject of multiple cases of legal action and risked bankruptcy and his reputation to expose misconduct in the pharmaceutical industry.
A very interesting article that’s worth reading. Fact is, that companies who have a strong economic interest in the scientific process will have an impact on the quality of the research. It is, again and again, horrible to learn how far companies try to go – and often successfully do. While medical companies has always been an obvious target (and perpetrator), the problem runs deeper than the narrative of “Big Pharma”.