PhD Researchers & Doctoral Candidates
A correctly specified panel, growth, or multilevel model for a thesis—with the within/between distinction handled properly.
Research Methods
Longitudinal, Panel & Multilevel Research
Data observed over time and nested in groups carries information a single snapshot never can—but only if the model respects its structure. We handle the dependence, the heterogeneity, and the dynamics correctly, so your within-unit effects and growth trajectories mean what you claim they mean.
Panel · longitudinal · multilevel
RI-CLPM · growth · continuous-time
R · Stata · Mplus
Reproducible, journal-ready
Sample output
average trajectory
individuals
Overview
Structure that a snapshot can't give you
Observing the same units repeatedly is what lets you separate a real within-unit effect from a stable difference between units—the distinction on which most longitudinal claims stand or fall. But that power comes with obligations: repeated observations are not independent, units differ in ways you can't always measure, and effects unfold over intervals your data may not have sampled evenly. Ignore any of these and the standard errors and conclusions are wrong.
We match the model to the structure. For firm- or country-level panels, fixed or random effects with the assumption tested rather than assumed; for nested data such as employees within firms or students within schools, multilevel and mixed-effects models that partition variance correctly; for reciprocal effects over time, the random-intercept cross-lagged panel model that isolates genuine within-person dynamics; for irregular waves, continuous-time models that don't depend on equal spacing.
Whether the deliverable is a panel-regression paper, a growth-trajectory study, or a multilevel model, the discipline is the same: the dependence handled, the right effects estimated, the diagnostics reported, and reproducible code you keep.
Who We Work With
For research that follows units over time
If your data is a panel, a set of waves, or nested in groups, this is the right desk to write to.
Economics & Finance Researchers
Firm-, country-, and bank-level panel studies to the standard the field's journals expect.
Psychology & Behavioral Scientists
Longitudinal designs—latent growth, RI-CLPM, dynamic SEM—for change and reciprocal effects over waves.
Education & Organizational Researchers
Multilevel and hierarchical models for students in schools, employees in firms, and other nested data.
Policy & Development Researchers
Household and regional panels, and small-area estimation, for evidence that tracks change over time.
Research Institutes & Data Teams
Groups managing long panel or survey-wave datasets who need the right longitudinal model applied.
Capabilities
The full longitudinal & multilevel toolkit
Organized by data type, dynamic model, and nesting structure. If your study needs a method not listed here, ask—this is the core, not the boundary.
Data observed over time
The panel structures we work across—from firm and country panels to household and financial-market data.
- Longitudinal studies
- Cross-sectional studies
- Panel studies
- Firm-level panels
- Country-level panels
- Household panels
- Regional panels
- City-level panels
- Industry panels
- Bank-level panels
- Financial-market panels
Dynamics & growth over time
Models for reciprocal effects, trajectories, and change measured across waves and continuous time.
- Cross-lagged panel models
- Random-intercept CLPM
- Latent growth models
- Growth mixture models
- Dynamic SEM
- Multilevel longitudinal models
- Time-varying effects
- Continuous-time models
- Continuous-time meta-analysis
Nested & grouped data
Models for data organized in levels—correctly partitioning variance and estimating cross-level effects.
- Multilevel modeling
- Hierarchical linear models
- Mixed-effects models
- Cross-level interactions
- Multilevel SEM
- Bayesian hierarchical models
- Small-area estimation
- Hierarchical longitudinal models
How the Analysis Works
Six steps from waves to within-unit effects
A transparent sequence that respects the structure of longitudinal data—dependence handled, the right effects separated, every choice documented. Nothing is a black box.
Steps are adapted to your data: many units vs. many waves, balanced vs. unbalanced, nested vs. flat, evenly vs. irregularly spaced. We confirm the model with you before estimation begins.
-
1
Structure
Map the data's structure—units, waves, nesting, and spacing—since it dictates which models are even valid.
Inputs: units · waves · nesting · balance
-
2
Specify
Choose the estimator or model—fixed/random effects, multilevel, growth, or cross-lagged—that matches the question and structure.
Models: FE/RE · multilevel · growth · RI-CLPM
-
3
Test assumptions
Run the assumption tests each model depends on and reason about whether they are credible for your data.
Tests: Hausman · ICC · serial correlation · heteroskedasticity
-
4
Estimate
Fit the model with correct standard errors for the panel or nested structure—clustered where dependence requires it.
Inference: clustered SE · robust · ML/REML · Bayesian
-
5
Stress-test
Check alternative specifications, sample splits, and attrition or missingness that could bias longitudinal results.
Checks: alternative specs · attrition · missingness · sensitivity
-
6
Report
Deliver trajectory and effect plots, tables, methodology, and reproducible code you keep.
Output: figures · tables · methods · R/Stata/Mplus code
Rigor by default
The checks that keep a longitudinal claim valid
The power of repeated measures comes with assumptions that reviewers scrutinize. Testing them is standard on every engagement.
Included on every project
- Fixed- vs. random-effects testing and justification
- Correct standard errors for panel or nested structure
- Within- vs. between-unit effects properly separated
- Attrition and missing-data assessment
- Reproducible, versioned code you keep
What You Receive
Every engagement, delivered in full
Not a black-box result and a number, but a complete, documented package you can submit, defend, and reproduce.
- Clean, structured panel or longitudinal dataset
- Model specification with assumption tests
- Estimation output with correct standard errors
- Within/between or growth decomposition
- Trajectory and effect plots
- Interpretation of effects across levels and time
- Reproducible R, Stata, or Mplus code
- Journal-ready methodology and results sections
- Technical responses to methodological reviewer comments, where required
Where this fits
Part of a larger arc
Longitudinal work is strongest when the design ahead of it anticipates the waves and the reporting after it is precise—each handled with the same care.
Research Design & Planning
Identification strategy, power, and specification decided before estimation begins.
Explore methodsStatistical & Methodological Audit
An independent check of assumptions, specification, and reproducibility before submission.
Explore auditPublication & Research Support
Methods and results reporting, journal selection, and reviewer-response support.
Explore support
FAQ
Common questions
Answers to what most researchers and project leads ask before we begin a panel or longitudinal engagement.
They overlap heavily. Economists tend to say panel data for many units observed over a few periods and use fixed- or random-effects estimators; psychologists and social scientists say longitudinal for fewer units observed over more waves and lean toward growth and cross-lagged models. We work across both traditions and choose the framework your question and data structure call for.
Fixed effects control for all stable unit characteristics and are the safer default when those may correlate with your predictors; random effects are more efficient and allow time-invariant regressors but require a stronger assumption. We run a Hausman test and, more importantly, reason about whether that assumption is credible for your data rather than relying on the test alone.
Yes—this is exactly what the random-intercept cross-lagged panel model (RI-CLPM) was designed for. The traditional cross-lagged panel model conflates the two; RI-CLPM and related models isolate genuine within-unit dynamics, and are increasingly required by reviewers for cross-lagged claims.
Yes. Nested data violates the independence assumption of ordinary regression, so we use multilevel, hierarchical, or mixed-effects models that correctly partition variance across levels and let you estimate cross-level interactions and random slopes.
Yes. Latent growth models estimate the average trajectory and individual variation around it; growth mixture models identify distinct latent classes of trajectories when the population is not homogeneous. We select based on your theory and test the number of classes rigorously.
It can be for discrete-time models, which assume equal intervals. Continuous-time models handle unequal spacing directly by estimating an underlying continuous process, giving parameters that are comparable regardless of the observation schedule—the right approach when waves are irregular.
R (plm, lme4, lavaan, and related packages), Stata, and Mplus, matched to the model and your environment. You receive versioned, commented code alongside the results and a methods section written to journal standards.
Yes, and it is the highest-leverage point. The number of waves, their spacing, and the sample size at each level determine which longitudinal models are even estimable—designing them in advance prevents a dataset that cannot answer the question it was collected for.
Related Expertise
Often paired with econometrics
Working with panel or longitudinal data?
Tell us your data's structure and your question—we'll tell you the right model, and make sure the within-unit effects hold up.