Statistical Methods in Theses: Guidelines and Explanations

Signed August 2018
Naseem Al-Aidroos, PhD,
Christopher Fiacconi, PhD
Deborah Powell, PhD,
Harvey Marmurek, PhD,
Ian Newby-Clark, PhD,
Jeffrey Spence, PhD,
David Stanley, PhD,
Lana Trick, PhD

Version: 2.00

This document is an organizational aid, and workbook, for students. We encourage students to take this document to meetings with their advisor and committee. This guide should enhance a committee’s ability to assess key areas of a student’s work. 

Context

In recent years a number of well-known and apparently well-established findings have failed to replicate, resulting in what is commonly referred to as the replication crisis. The APA Publication Manual 6thEdition notes that “The essence of the scientific method involves observations that can be repeated and verified by others.” (p. 12). However, a systematic investigation of the replicability of psychology findings published in Science revealed that over half of psychology findings do not replicate (see a related commentary in Nature). Even more disturbing, a Bayesian reanalysis of the reproducibility project showed that 64% of studies had sample sizes so small that strong evidence for or against the null or alternative hypotheses did not exist. Indeed, Morey and Lakens (2016) concluded that most of psychology is statistically unfalsifiable due to small sample sizes and correspondingly low power (see article). Our discipline’s reputation is suffering. News of the replication crisis has reached the popular press (e.g., The Atlantic,  The Economist,  Slate,Last Week Tonight).

An increasing number of psychologists have responded by promoting new research standards that involve open science and the elimination of Questionable Research Practices. The open science perspective is made manifest in the Transparency and Openness Promotion (TOP) guidelines for journal publications. These guidelines were adopted some time ago by the Association for Psychological Science. More recently, the guidelines were adopted by American Psychological Association journals (see details) and journals published by Elsevier (see details). It appears likely that, in the very near future, most journals in psychology will be using an open science approach. We strongly advise readers to take a moment to inspect the TOP Guidelines Summary Table

A key aspect of open science and the TOP guidelines is the sharing of data associated with published research (with respect to medical research, see point #35 in the World Medical Association Declaration of Helsinki). This practice is viewed widely as highly important. Indeed, open science is recommended by all G7 science ministers. All Tri-Agency grants must include a data-management plan that includes plans for sharing: “research data resulting from agency funding should normally be preserved in a publicly accessible, secure and curated repository or other platform for discovery and reuse by others.” Moreover, a 2017 editorial published in the New England Journal of Medicineannounced that the International Committee of Medical JournalEditorsbelieves there is “an ethical obligation to responsibly share data.” As of this writing, 60% of highly ranked psychology journals require or encourage data sharing.

The increasing importance of demonstrating that findings are replicable is reflected in calls to make replication a requirement for the promotion of faculty (see details in Nature) and experts in open science are now refereeing applications for tenure and promotion (see details at the Center for Open Science and this article). Most dramatically, in one instance, a paper resulting from a dissertation was retracted due to misleading findings attributable to Questionable Research Practices. Subsequent to the retraction, the Ohio State University’s Board of Trustees unanimously revoked the PhD of the graduate student who wrote the dissertation (see details). Thus, the academic environment is changing and it is important to work toward using new best practices in lieu of older practices—many of which are synonymous with Questionable Research Practices. Doing so should help you avoid later career regrets and subsequent public mea culpas. One way to achieve your research objectives in this new academic environment is to incorporate replications into your research. Replications are becoming more common and there are even websites dedicated to helping students conduct replications (e.g., Psychology Science Accelerator) and indexing the success of replications (e.g.,Curate Science). You might even consider conducting a replication for your thesis (subject to committee approval).

As early-career researchers, it is important to be aware of the changing academic environment. Senior principal investigators may be reluctant to engage in open science (see this student perspective in a blog post and podcast) and research on resistance to data sharing indicates that one of the barriers to sharing data is that researchers do not feel that they have knowledge of how to share data online. This document is an educational aid and resource to provide students with introductory knowledge of how to participate in open science and online data sharing to start their education on these subjects. 

Guidelines and Explanations

In light of the changes in psychology, faculty members who teach statistics/methods have reviewed the literature and generated this guide for graduate students. The guide is intended to enhance the quality of student theses by facilitating their engagement in open and transparent research practices and by helping them avoid Questionable Research Practices, many of which are now deemed unethical and covered in the ethics section of textbooks.

This document is an informational tool.

How to Start

In order to follow best practices, some first steps need to be followed. Here is a list of things to do:

  1. Get an Open Science account. Registration at osf.io is easy!
  2. If conducting confirmatory hypothesis testing for your thesis, pre-register your hypotheses (see Section 1-Hypothesizing). The Open Science Foundation website has helpful tutorials and guides to get you going.
  3. Also, pre-register your data analysis plan. Pre-registration typically includes how and when you will stop collecting data, how you will deal with violations of statistical assumptions and points of influence (“outliers”), the specific measures you will use, and the analyses you will use to test each hypothesis, possibly including the analysis script. Again, there is a lot of help available for this. 

Exploratory and Confirmatory Research Are Both of Value, But Do Not Confuse the Two

We note that this document largely concerns confirmatory research (i.e., testing hypotheses). We by no means intend to devalue exploratory research. Indeed, it is one of the primary ways that hypotheses are generated for (possible) confirmation. Instead, we emphasize that it is important that you clearly indicate what of your research is exploratory and what is confirmatory. Be clear in your writing and in your preregistration plan. You should explicitly indicate which of your analyses are exploratory and which are confirmatory. Please note also that if you are engaged in exploratory research, then Null Hypothesis Significance Testing (NHST) should probably be avoided (see rationale in Gigerenzer (2004) and Wagenmakers et al., (2012)). 

This document is structured around the stages of thesis work: hypothesizing, design, data collection, analyses, and reporting – consistent with the headings used by Wicherts et al. (2016).We also list the Questionable Research Practices associated with each stage and provide suggestions for avoiding them. We strongly advise going through all of these sections during thesis/dissertation proposal meetings because a priori decisions need to be made prior to data collection (including analysis decisions). 

To help to ensure that the student has informed the committee about key decisions at each stage, there are check boxes at the end of each section.

How to Use This Document in a Proposal Meeting

  1. Print off a copy of this document and take it to the proposal meeting.
  2. During the meeting, use the document to seek assistance from faculty to address potential problems.
  3. Revisit responses to issues raised by this document (especially the Analysis and Reporting Stages) when you are seeking approval to proceed to defense.

Consultation and Help Line

Note that the Center for Open Science now has a help line (for individual researchers and labs) you can call for help with open science issues. They also have training workshops. Please see their website for details.