PhD Researchers & Doctoral Candidates
A properly powered, cleanly identified experiment for a thesis—designed before launch and analyzed correctly after.
Research Methods
Experimental & Behavioural Research
An experiment is the cleanest way to establish a causal effect—but only if the design is right before it runs. We build randomization, treatment arms, and power into the study from the start, so the difference you measure between conditions is the effect you set out to test, not an artifact.
RCTs · lab · field · survey
Choice experiments & conjoint
oTree · Qualtrics · R
Reproducible, journal-ready
Sample output
control
treatment arms
Overview
The effect is decided at the design stage
Randomization is what makes an experiment special: assign treatment at random and, on average, the groups differ only in the treatment, so a difference in outcomes is a causal effect. But that logic is fragile. An underpowered study can't detect the effect it was built to find; a confounded manipulation muddies what the treatment even was; a flawed randomization quietly reintroduces the bias the design was meant to remove. None of these can be fixed after the data is collected.
So the work that matters most happens before launch. We build the randomization scheme, define treatment arms and controls that isolate the mechanism you care about, and run a formal power analysis so the sample is large enough—and no larger. Where it strengthens the study, we help pre-register the design and analysis plan, which is increasingly what top journals expect.
Then we analyze it correctly: the right test for the design, adjustment for multiple arms, attention to attrition and non-compliance, and—for choice experiments—the discrete-choice and conjoint models that recover preferences. The deliverable is a clean causal estimate with reproducible code you keep.
Who We Work With
For research that tests, not just observes
If you want to manipulate a condition and measure its effect, this is the right desk to write to.
Behavioural Economists & Scientists
Lab and online experiments testing biases, preferences, and decision-making with rigorous designs.
Marketing & Consumer Researchers
Choice experiments, conjoint studies, and consumer experiments that recover preferences and willingness-to-pay.
Finance & Investor-Behaviour Researchers
Behavioural-finance experiments on risk, framing, and investor decision-making.
Policy & Development Teams
Field experiments and RCTs evaluating interventions where causal evidence is the standard.
Survey & Applied Researchers
Survey experiments—vignette, framing, and priming designs—embedded in larger data collections.
Capabilities
The full experimental toolkit
Organized by experiment type, preference-elicitation method, and behavioural field. If your study needs a design not listed here, ask—this is the core, not the boundary.
Control vs. realism
The full range of experimental designs, from tightly controlled lab studies to real-world field trials.
- Randomized controlled trials
- Laboratory experiments
- Field experiments
- Behavioural experiments
- Survey experiments
Choice & conjoint methods
Designs that recover how people value attributes and trade them off against each other.
- Choice experiments
- Discrete choice experiments
- Conjoint analysis
- Consumer experiments
Decisions & markets
The behavioural domains where experimental methods are most often applied.
- Investor behaviour research
- Behavioural economics
- Behavioural finance
How the Analysis Works
Six steps from hypothesis to clean effect
A transparent sequence weighted toward the front—because in experiments, most of the value is created before any data is collected. Nothing is a black box.
Steps are adapted to your design: lab vs. field vs. survey, between- vs. within-subject, single- vs. multi-arm. We settle the design and power with you before launch.
-
1
Design
Translate the hypothesis into treatment arms, controls, and a randomization scheme that isolates the mechanism.
Inputs: hypothesis · arms · controls · randomization
-
2
Power
Run a formal power analysis to size the sample for the effect you need to detect, accounting for arms and clustering.
Analysis: power · MDE · multiple arms · ICC
-
3
Build & pre-register
Program the instrument, pilot it, and—where apt—pre-register the design and analysis plan.
Tools: oTree · Qualtrics · Gorilla · pre-registration
-
4
Run
Field the experiment with randomization checks and monitoring for attrition and data quality.
Checks: balance · attrition · manipulation checks
-
5
Analyze
Estimate treatment effects with the correct test, adjust for multiple comparisons, and handle non-compliance.
Methods: ANOVA/regression · ITT/LATE · choice models
-
6
Report
Deliver treatment-effect figures, tables, methodology, and reproducible instrument and analysis code you keep.
Output: figures · tables · methods · instrument & R code
Rigor by default
The checks that make an experiment credible
A treatment effect is only believable if the design supports it. The safeguards that protect an experiment are standard on every engagement.
Included on every project
- Formal power analysis and sample-size justification
- Randomization and balance checks across arms
- Manipulation and attention checks
- Attrition, non-compliance, and multiple-comparison handling
- Reproducible instrument and analysis 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.
- Experimental design and randomization scheme
- Power analysis and sample-size justification
- Programmed experimental instrument
- Balance and manipulation-check results
- Treatment-effect estimates with figures
- Interpretation of effects and heterogeneity
- Reproducible R or Stata analysis code
- Journal-ready methodology and results sections
- Technical responses to methodological reviewer comments, where required
Where this fits
Part of a larger arc
An experiment is strongest when the design ahead of it is deliberate 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 an experimental or behavioural engagement.
Yes, and design is where an experiment succeeds or fails. Randomization scheme, treatment arms, control conditions, and power all have to be settled before a single participant is run—a beautifully analyzed but underpowered or confounded experiment cannot be rescued after the fact. We help build the design first, then analyze it.
Through a formal power analysis tied to the effect size you care about detecting, the number of arms, and your significance and power targets. For multi-arm or clustered designs we account for multiple comparisons and intra-cluster correlation, so the study is neither underpowered nor wastefully large.
They trade off control against realism. Lab experiments maximize control but may not generalize; field experiments maximize realism but are harder to run cleanly; survey experiments (including vignette and framing designs) sit in between and scale easily. We recommend the type that best matches your question and constraints, and are candid about each one's limits.
Yes. We design efficient choice sets, run discrete choice experiments and conjoint studies, and estimate the models—conditional and mixed logit, hierarchical Bayes—that recover attribute preferences and willingness-to-pay, common in marketing, health, and environmental economics.
Where appropriate, yes—and we recommend it. A pre-registered design and analysis plan protects your study from the criticism that hypotheses or tests were chosen after seeing the data, which matters increasingly for publication in top journals.
Yes. We support behavioural research—biases, heuristics, framing, risk and time preferences, investor behaviour—using experimental and observational designs suited to testing behavioural hypotheses rigorously rather than illustrating them anecdotally.
For running experiments, platforms such as oTree, Qualtrics, and Gorilla; for analysis, R and Stata. You receive the experimental instrument, the randomization and analysis code, and a methods section written to journal standards.
Yes, and it is the highest-leverage point. Reviewing the design, power, randomization, and pre-analysis plan before launch prevents the fatal flaws—confounds, insufficient power, ambiguous manipulations—that no amount of later analysis can fix.
Related Expertise
Often paired with econometrics
Planning an experiment?
Tell us your hypothesis and constraints—we'll help design it, power it, and analyze it so the effect you measure is the one you set out to test.