On social media, there are often discussions about why psychologists leave academia. Some argue that the new culture of criticism (where overly harsh criticism is leveled against those who make errors and accusations of malfeasance are rife) make academics, especially younger ones, change profession. They give examples of high profile cases where researchers have made errors or merely used previously accepted standards of practice and the obloquy they’ve received when their results don’t hold up. Others provide counter examples of former colleagues who grew frustrated at their inability to replicate supposedly rock-solid findings or who had a crisis in confidence about the validity of vast swathes of the literature. But one thing is lacking from this discussion.
A few weeks ago, Nature published an article summarising the various measures and counter-measures suggested to improve statistical inferences and science as a whole (Chawla, 2017). It detailed the initial call to lower the significance threshold to 0.005 from 0.05 (Benjamin et al., 2017) and the paper published in response (Lakens et al., 2017). It was a well written article, with one minor mistake: an incorrect definition of a p-value:
The two best sources for the correct definition of a p-value (along with its implications and examples of how a p-value can be misinterpreted) are Wasserstein & Lazar (2016) and its supplementary paper Greenland et al. (2016). A p-value has been defined as: “a statistical summary of the compatibility between the observed data and what we would predict or expect to see if we knew the entire statistical model (all the assumptions used to compute the P value) were correct” (Greenland et al., 2016). To put it another way, it tells us the probability of finding the data you have or more extreme data assuming the null hypothesis (along with all the other assumptions about randomness in sampling, treatment, assignment, loss, and missingness, the study protocol, etc.) are true. The definition provided in the Chawla article is incorrect because it states “the smaller the p-value, the less likely it is that the results are due to chance”. This gets things backwards: the p-value is a probability deduced from a set of assumptions e.g. the null hypothesis is true, so it can’t also tell you the probability of that assumption at the same time. Joachim Vandekerckhove and Ken Rothman give further evidence as to why this definition is incorrect:
Anyone who has taken one of the harder sciences at university or knows someone who has will know what “labs” are. You are given practical assignments to complete that are meant to consolidate what you’ve learnt in the lecture/seminar. They are almost ubiquitous in physics after becoming widespread by the beginning of the 20th century (Meltzer & Otero, 2015), as they are for chemistry (Layton, 1990), and biology (Brownell, Kloser, Fukami, & Shavelson, 2012). Their value is widely assumed to have been demonstrated multiple times across the hard sciences (Finkelstein et al., 2005; Blosser, 1990) but questions have occasionally been raised as to their effectiveness (Hofstein & Lunetta, 2004). A new paper by Holmes, Olsen, Thomas, & Wieman (2017) sought to test whether participating in labs actually improved physics students’ final grades or not. Across three American universities they tested three questions: what is the impact of labs on associated exam performance; did labs selectively impact the learning physics concepts; and are there short-term learning benefits that are “washed out on the final exams”?
Every once in a while in the psychology sphere of social media there’s a discussion about why people leave academia. This talking point often comes up in the context of “the open science movement” and whether more academics leave because of the cultural of criticism or because of the lack of replicability of some findings. People who have left academia offer their reasons and people who are still in give several anecdotes about why someone they know left. But what seems to be lacking is some actual data. So I’ve written this survey with the hopes of shedding some light on the situation. It’s for people who have considered leaving or have actually left academia or practicing psychology (educational, clinical, etc.). But this survey will only be useful if you share this with people you know who have left. So please share the survey on social media or relevant mailing lists but especially link it directly to people you know who have left psychology. I’m writing this blog post so those who are subscribed to the methods blog feed will see this survey, hopefully increasing the number of respondents. Thank you for your help.
I’m a big fan of Twitter and have learned so much from the people on there, so I’m always happy to share someone singing it’s praises. This article was written by Jean-Jacques Orban de Xivry for the University of Leuven’s blog. He talks about how he uses it to find out about interesting papers and a whole host of other benefits. The article can be found here. His twitter account is: @jjodx.
Do you think you know which studies will replicate? Do you want to help improve the replicability of science? Do you want to make some money? Then take part in this study on predicting replications!
But why are researchers gambling on whether a study will successfully replicate (defined as finding “an effect in the same direction as the original study and a p-value<0.05 in a two-sided test” for this study)? Because there is some evidence to suggest that a prediction market can be a good predictor of replicability, even better than individual psychology researchers.
There has been a lot of tension in the psychological community recently. Replications are becoming more prevalent and many of them are finding much smaller effects or none at all. This then raises a lot of uncomfortable questions: is the studied effect real? How was it achieved in the first place? Were less than honest methods used (p-hacking etc.)? The original researchers can sometimes feel that these questions go beyond valid criticisms to full-blown attacks on their integrity and/or their abilities as a scientist. This has led to heated exchanges and some choice pejoratives being thrown about by both “sides”.
The Reproducibility Project: Psychology:
The Reproducibility Project: Psychology (OSC, 2015) was a huge effort by many different psychologists across the world to try and assess whether the effects of a selection of papers could be replicated. This was in response to the growing concern about the (lack of) reproducibility of many psychological findings with some high profile failed replications being reported (Hagger & Chatzisarantis, 2016 for ego-depletion and Ranehill, Dreber, Johannesson, Leiberg, Sul, & Weber, 2015 for power-posing). They reported that of the 100 replication attempts, only ~35 were successful. This provoked a strong reaction not only in the psychological literature but also in the popular press, with many news outlets reporting on it.
This post is a response to “Pre-Registration of Analysis of Experiments is Dangerous for Science” by Mel Slater (2016). Preregistration is stating what you’re going to do and how you’re going to do it before you collect data (for more detail, read this). Slater gives a few examples of hypothetical (but highly plausible) experiments and explains why preregistering the analyses of the studies (not preregistration of the studies themselves) would not have worked. I will reply to his comments and attempt to show why he is wrong.