Evidence for your
dissertation decisions.

Across 5,876 clinical psychology dissertation and thesis abstracts since 2021, mediation, moderation, and regression appear more often than anywhere else in psychology. Here is the full picture, and what it means for your own study.

In 6,105 social psychology dissertation and thesis abstracts since 2021, mediation and moderation are unusually common, dyadic models stand out, and reliability reporting has doubled.

Developmental and educational psychology dissertations name t tests three times as often as the rest of psychology and reliability twice as often, while multilevel models remain rare. A look at 5,266 abstracts since 2021.

Experimental and cognitive psychology dissertations name machine learning more than three times as often as the rest of psychology, along with network analysis and Bayesian methods. A look at 3,354 abstracts since 2021.

Applied psychology is the smallest subfield in our data, and the most intervention-focused: 707 abstracts since 2021, with the highest share of experiments and quantitative methods in psychology.

Across 4,624 abstracts on teaching, learning, and curriculum since 2021, the pretest–posttest toolkit dominates: t tests, reliability, and ANOVA, with little regression or modeling.

Online and technology-enhanced learning dissertations name t tests in 44% of method-naming abstracts, the highest share of any area we have analyzed. A look at 2,487 abstracts since 2021.

Higher and adult education dissertations are the most qualitative in education, and their quantitative work leans on regression, structural equation models, and student outcomes like retention. A look at 1,727 abstracts since 2021.

Educational leadership, teacher, and policy dissertations use regression, mediation, and causal designs far more than the rest of education, and structural equation models are rising. A look at 2,699 abstracts since 2021.

Early childhood, family, and inclusive education dissertations borrow from developmental psychology: mediation, moderation, parent–child dyads, and latent profiles. A look at 2,016 abstracts since 2021.

Strategy and management dissertations combine firm-level panel data with moderation-heavy theory testing. A look at 5,124 abstracts since 2021.

Accounting and finance dissertations are the most quantitative in business, built on regression, panel models, causal designs, and time series. A look at 6,650 abstracts since 2021.

Marketing dissertations name structural equation models more than any other method: SEM and PLS-SEM each appear in 24% of method-naming abstracts. A look at 3,867 abstracts since 2021.

Organizational behavior and HRM dissertations name mediation in 28% of method-naming abstracts, more than any other area we have analyzed. A look at 2,672 abstracts since 2021.

Information systems and technology management dissertations name machine learning more than any other method, alongside survey models of technology adoption. A look at 3,866 abstracts since 2021.

Sociology is the largest field in our data and one of the least quantitative. When sociology abstracts name a method, mediation and moderation stand out. A look at 22,999 abstracts since 2021.

Political science and international relations is the least quantitative field in our data: only 12% of abstracts name a statistical method. Those that do lean on regression, panel data, and text and network methods. A look at 9,301 abstracts since 2021.

Economics dissertations are the most quantitative in the social sciences, built on panel data, time series, and causal designs, and all three are growing. A look at 7,087 abstracts since 2021.

Public health dissertations name logistic regression twice as often as the other social sciences, along with survival and count models, and they have the most trials and systematic reviews. A look at 3,778 abstracts since 2021.

Nursing and health professions dissertations are survey-heavy and outcome-focused: logistic regression, chi-square, correlation, and reliability lead. A look at 5,636 abstracts since 2021.

Communication is the most qualitative field in this series, but when its abstracts name a method, mediation leads, and text and network analysis are unusually common. A look at 1,998 abstracts since 2021.

The results chapter reports what you found, question by question, with the numbers a reader needs to judge it. Here is what APA 7 and its reporting standards ask for, how to format the statistics, and example write-ups you can adapt.

Mediation is one of the most common analyses in psychology, education, and business dissertations, and one of the least completely reported. Here is how to run Hayes PROCESS Model 4, what each part of the output means, and how to write it up so your committee can check it.

Structural equation modeling lets you test relationships between constructs while accounting for how they were measured. It also gives a committee more to question than almost any other method. Here is how to plan it, run it, judge its fit, and report it.

Almost every survey-based dissertation has to show that its scales measure what they claim to. Factor analysis is how. Here is when to use exploratory or confirmatory factor analysis, the decisions each one involves, and how to report them.

Students in schools, employees in teams, patients in clinics, and days within people: when your data are nested, ordinary regression gives standard errors that are too small. Here is how to build a multilevel model step by step and report it so your committee can follow it.

When you can’t randomize, propensity scores let you compare people who received a program with similar people who didn’t. They can’t fix confounders you didn’t measure, and committees know it. Here is how to do it carefully and report it honestly.

Latent profile and latent class analysis find subgroups of people who share a pattern of responses. Choosing how many subgroups there are is the hardest part, and the part committees question most. Here is how to do it defensibly and report it clearly.

Nearly every dissertation with a survey or a coding scheme reports reliability, usually as a single Cronbach’s alpha. Committees increasingly ask for more. Here is how to choose the right coefficient, what alpha can and can’t tell you, and how to report reliability properly.