Research Methods

Fifteen methods, each a full capability

A research method defines how a question is answered. We work across fifteen method areas—from econometrics to qualitative research—and each is a genuine specialism, matched to the question and the data rather than applied by default. Every method pairs with one of our six subject domains on every project.

Fifteen method areas Matched to your question Genuine specialists
Research methods Fifteen method areas represented as a cluster of connected nodes. methods · fifteen areasfifteen specialisms, one standard
Research methods methods
Overview

The method is a decision, not a default

A research method is the machinery that turns a question and a dataset into a defensible answer. It determines what your study can credibly claim, which assumptions it rests on, and whether the result will survive the scrutiny of a referee who does this for a living. Yet in practice the method is too often chosen by habit rather than by fit—the technique the researcher already knows, or the one a supervisor happens to favour, applied to a question it was never designed for. That mismatch is behind a large share of avoidable rejections: a pooled regression run on data that is really a panel, a mediation model estimated with an approach reviewers now reject, a forecast judged on how well it fits the past rather than how well it predicts the future.

We work the other way around. We start from the question you are actually asking and the structure of the data you actually have, and only then choose the method that fits both. Sometimes that points to a workhorse technique done rigorously; sometimes to a more specialized estimator that the question genuinely requires. The point is that the choice is deliberate and justified, made explicit so that a reviewer can follow the reasoning as easily as you can. Across fifteen method areas—spanning quantitative and econometric analysis, modelling and machine learning, and synthesis, primary, and qualitative research—we bring genuine specialists rather than one generalist stretched thin, because for anything non-trivial, depth in the specific method beats broad familiarity with all of them.

Each of these fifteen areas is a full capability in its own right, with its own literature, its own standards, and its own pitfalls. And each pairs, on every project, with one of our six subject domains—the field you are studying and the analytical approach you are using, brought together by people who know both. What follows is a map of the fifteen, grouped into three families so you can see how they relate; but the grouping is a convenience, not a boundary, and many of the strongest studies draw on more than one.

Our method

How we choose the right method

Choosing a method well is a structured judgement, not a lookup. Three questions, in order, narrow the field to a small, defensible set—before any estimator is chosen.

01

What claim are you making?

A descriptive question—what is the pattern, how has it changed, how do things cluster—calls for very different tools than a causal question about whether one thing produced another, which in turn differs from a predictive question about what happens next. Confusing these is a common, costly error: a model built to predict will not credibly establish cause, and a model built to estimate an effect is usually a poor forecaster. We settle the nature of the claim first, because it narrows the field immediately.

02

What is the structure of the data?

Is it a cross-section, a time series, or a panel across units and time? Are observations independent or nested—students within schools, repeated measures within people? Is the outcome continuous, binary, a count, a duration? Each feature rules methods in or out, and getting it wrong produces standard errors that are quietly incorrect. A short dynamic panel points toward a dynamic-panel estimator, not ordinary fixed effects; non-stationary macro series of mixed order point toward bounds-testing, not a naive VAR.

03

What does the field expect?

Methods have conventions, and those conventions move. A technique standard five years ago may now draw a rejection—the two-way fixed-effects difference-in-differences estimator, biased under staggered adoption; the causal-steps approach to mediation, superseded by bias-corrected bootstrapping; the cross-lagged panel model, now expected to separate within- from between-person change. Knowing what current reviewers accept is exactly the specialized, up-to-date knowledge our specialists bring—and once all three questions are settled, we select the design and document the reasoning so it can be defended.

The fifteen methods

Three families, fifteen full capabilities

The grouping below organizes the fifteen areas by what they do—estimating relationships, modelling and predicting, or synthesizing and gathering evidence. Each links to a full page describing exactly how that work is done.

Quantitative & Econometric

The estimation core—models for economic, financial, and firm data.

Econometrics & Quantitative Research

Classical, time-series, and advanced panel econometrics with full diagnostics.

  • OLS & GLS
  • Fixed & random effects
  • IV / 2SLS
  • GMM & dynamic panels
  • Quantile regression
  • Limited dependent variables
  • Missing-data & imputation
  • Diagnostics & robustness
Explore area

Causal Inference & Policy Evaluation

DiD, RDD, IV, matching, and synthetic control—identification tested, not assumed.

  • Difference-in-differences
  • Regression discontinuity
  • Instrumental variables
  • Matching & PSM
  • Synthetic control
  • Propensity scores
  • Quasi-experimental design
Explore area

Financial Econometrics & Risk Analytics

GARCH volatility, spillovers, wavelets, event studies, and risk modelling.

  • GARCH family
  • Volatility modelling
  • Spillover & connectedness
  • Wavelet analysis
  • Event studies
  • Value-at-Risk
  • Asset pricing
Explore area

Forecasting & Predictive Analytics

ARIMA, VAR, and ML forecasting, evaluated honestly out-of-sample.

  • ARIMA / SARIMA
  • VAR / VECM
  • State-space models
  • Structural time series
  • Bayesian time series
  • Regime-switching
  • MIDAS
  • Out-of-sample validation
Explore area

Longitudinal, Panel & Multilevel Research

Panel, growth, survival, and hierarchical models for data over time.

  • Panel & dynamic panels
  • Multilevel / mixed models
  • Latent growth
  • RI-CLPM
  • Survival / Cox regression
  • Kaplan–Meier
  • Competing risks & frailty
  • Repeated measures
Explore area

Spatial, Regional & Geospatial Analytics

Spatial econometrics, GIS, GWR, and regional convergence.

  • Spatial econometrics (SAR/SEM)
  • Geographically weighted regression
  • Moran's I / LISA
  • Spatial clustering
  • GIS & geospatial statistics
  • Spatial interpolation
  • Regional convergence
Explore area
Synthesis, Primary & Qualitative

Reviewing evidence, collecting data, and studying meaning.

Depth over breadth

Why genuine method depth matters

Methods are not interchangeable tools any competent analyst can pick up as needed. The difference between an analysis that merely runs and one that persuades lives in the details only a specialist attends to.

Each of these fifteen areas has a deep literature, standards that shift over time, and subtle pitfalls that separate a defensible result from a fragile one. That difference is rarely visible in the headline output—the coefficient, the p-value, the model fit—and almost always visible in the details: whether the identifying assumption was tested rather than assumed, whether the measurement was validated before the structural model was estimated, whether the forecast was checked on data the model had never seen, whether the robustness section anticipates the specific objection a reviewer will raise.

A generalist stretched across every method inevitably works at the level of someone who learned each from a textbook—precisely the level at which these subtle failures go unnoticed. The specialist, by contrast, uses the method routinely, follows its evolving literature, and knows where its bodies are buried. This is why we are organized as a network of specialists: when your project needs a random-intercept cross-lagged panel model, a modern staggered difference-in-differences estimator, or a fuzzy-set qualitative comparative analysis, it is handled by someone who does that method as a matter of course.

That depth pays off twice. First, in quality: the analysis is done correctly, with the diagnostics and safeguards the method actually requires, so the result holds up under its own stress tests rather than collapsing at the first hard question. Second, in credibility: a study that applies its method to current standards, in the register its field expects, reads as the work of someone who belongs in that conversation—which is half the battle in getting past a skeptical referee.

Depth is not a luxury or a premium add-on in methodological work; it is the difference between a result that survives review and one that does not. It is also why the fifteen flagship areas are genuine specialisms rather than labels on a menu: behind each is a person who could referee for the journals you are writing into, not a generalist meeting the method for the first time on your data and your deadline.

Methods in combination

The strongest studies rarely use just one

The three families are a way of organizing the map, not walls between the territories. Real research questions routinely reach across them, and part of our value is bringing the right combination together, coordinated rather than stitched.

Estimate, then stress-test

A causal-inference study establishes an effect; a robustness and sensitivity pass, and often an independent audit, decides whether it holds. Estimation and validation are different capabilities applied to the same result.

Measure, then model

A management study using SEM depends first on psychometrics—valid, reliable measurement of its constructs—before any structural path is estimated. Get the measurement wrong and the model, however sophisticated, is meaningless.

Map, then synthesize

A review begins with bibliometric mapping to see the shape of a field, then meta-analysis to quantify what its studies collectively show. Description and synthesis work in sequence.

Predict, then explain

A machine-learning model can predict an outcome accurately, but interpretable-ML and causal methods are what turn a prediction into an explanation a decision-maker can act on.

Qualitative, then quantitative

A mixed-methods design uses qualitative work to build a construct or hypothesis and quantitative methods to test it at scale—each method doing what it does best, integrated by design.

Model, then decide

An economic model projects outcomes under scenarios; operations-research and decision methods turn those projections into a structured, defensible choice among options.

Start from the question

Research questions we answer

If you are not sure which method you need, start here. The question you are asking points directly to the family of methods that can answer it credibly.

Does X cause Y?

Establishing causation from observational or experimental data.

Causal inference · experimental design · policy evaluation · propensity scores

What will happen next?

Forecasting an outcome and quantifying how confident you can be.

Forecasting · time-series · predictive ML · Bayesian methods

What explains this relationship?

Estimating and interpreting how variables relate, with correct inference.

Econometrics · regression · SEM · multilevel models

How does this change over time?

Analysing trajectories, growth, and time-to-event in longitudinal data.

Panel & longitudinal · survival analysis · growth models · time series

What is the difference between groups?

Comparing outcomes across groups or conditions, with adequate power.

Experimental design · ANOVA · power & sample size · survey methods

What does the evidence collectively show?

Synthesising many studies into an authoritative, quantified conclusion.

Systematic review · meta-analysis · network meta-analysis · bibliometrics

How do I measure this construct?

Building and validating a reliable, valid measurement instrument.

Psychometrics · IRT / Rasch · factor analysis · scale development

What is the best decision?

Choosing optimally among options under multiple criteria or constraints.

Operations research · MCDM · optimization · decision analysis

How do people or things behave?

Understanding meaning, process, and behaviour in depth.

Qualitative · mixed methods · behavioural research · text-as-data

Across every method

Capabilities that run across the methods

Some capabilities are not a single method area but run through all of them—the design that comes before analysis, the standards that run through it, and the domains it is applied in. These are part of every engagement, whichever flagship method it uses.

Before the analysis

Research design & methodology

Study design, hypothesis and framework development, variable operationalization, identification strategy, and sampling—the decisions that determine whether a study can answer its question at all.

Design essentials

Power & sample size

Sample-size determination and statistical power analysis for regression, ANOVA, SEM, longitudinal, and experimental designs—including simulation-based power where the design is complex.

Data foundations

Data quality & missing data

Missing-data analysis (MCAR/MAR/MNAR), multiple imputation, outlier and measurement-error handling, and selection-bias assessment—so results rest on data that can bear them.

Health economics & policy

Health economics & outcomes research

Applied economic and quantitative support for health-economics, health-policy, and healthcare-management research—economic evaluation, cost-effectiveness and cost-benefit analysis, budget-impact analysis, health-outcomes research, health-policy evaluation, and the econometric and causal methods that underpin them.

Statistical computing

Reproducible computational workflows

Analysis delivered in R, Python, Stata, and specialist packages, with versioned, documented, reproducible pipelines—so every result can be re-run from the raw data.

Standards

Reproducibility & research integrity

Reproducible statistical analysis, documented data and code, analysis audit, and replication—the research-transparency standards that increasingly separate credible work from the rest.

Looking for a capability by name—panel econometrics, causal inference, SEM, power analysis, Bayesian modelling, network analysis, or evidence synthesis? Each lives within one of the fifteen flagship areas above; tell us your research question and we'll point you to the right one.

Common questions

Choosing and combining methods

What researchers most often ask when they are not sure which method their project needs.

Yes, and it is one of the most valuable things we do. Tell us the question you are asking and the data you have or plan to collect, and we will recommend the method that fits both—and explain why, including the alternatives we considered. This advice is best sought early, before you commit time to an analysis that might answer the wrong question.
Three things, in order: the kind of claim you want to make (descriptive, causal, or predictive), the structure of your data (cross-section, time-series, or panel; independent or nested; the type of outcome), and what your field's reviewers currently expect. Together these usually narrow the appropriate methods to a small, defensible set.
Very often it should. Strong studies routinely combine methods—measurement before modelling, estimation before validation, mapping before synthesis, prediction before explanation. We bring the relevant specialists together and coordinate them, so the combination is coherent rather than a patchwork.
These fifteen are the core, not the boundary. They cover the overwhelming majority of quantitative, qualitative, and mixed-methods research in management and allied studies, but if your project needs a method not listed here, ask—we will tell you honestly whether we can help or point you toward someone better placed.
Every project pairs a method with a subject domain—the analytical approach and the field being studied. The strength of our work comes from bringing genuine expertise in both to the same engagement, so the person applying the method also knows your literature, your debates, and your journals.
That is common, and it is exactly where a current specialist helps. Several standard methods have been revised in recent years in ways reviewers now expect—in difference-in-differences, mediation analysis, and cross-lagged panel models, among others. We apply methods to their current standards, so your analysis does not draw an avoidable objection for using a superseded approach.

Not sure which method fits?

Tell us your question and your data—we'll recommend the right method and explain why.