Can you test whether a new teaching method works if you cannot randomly assign students to classes? You can, with a quasi-experimental design, and many master’s theses in education, nursing, management and public health do exactly that. The price is weaker control over alternative explanations, so the design and the write-up must show how you dealt with them. This guide explains true experiments and quasi-experiments, six designs that suit a thesis, the main threats to validity, the statistics that match each design, and how to report them.
This article is part of our guide How to Write a Master’s Thesis by the Mastermind PhD team and relates to our Statistical Analysis service. Sources last checked: November 2026.
What is a quasi-experimental design? A quasi-experimental design is a study that manipulates or compares an intervention and measures its effect on an outcome, like an experiment, but without random assignment of participants to conditions. Groups are formed by existing units (classes, wards, branches) or by time, so the researcher must use pretests, comparison groups or repeated measurements to rule out rival explanations.
What makes a study a true experiment?
Shadish, Cook and Campbell, whose 2002 book is the standard reference, define a randomized experiment by three features:
- Manipulation: the researcher introduces the treatment or intervention.
- Random assignment: chance alone decides who receives which condition.
- Control: a comparison condition and standardized procedures.
Random assignment makes the groups equivalent on average on every variable, measured or not, so a difference in the outcome can be attributed to the treatment. Randomized controlled trials in medicine follow this logic and are reported with the CONSORT 2010 statement.
How does a quasi-experimental design differ?
A quasi-experimental design keeps manipulation and comparison but drops random assignment. That is often unavoidable in a thesis: a school principal assigns whole classes, a hospital introduces a protocol on one ward, a company rolls out training in one branch. Because groups may differ before the intervention, the researcher must measure and adjust for those differences.
| Feature | True experiment | Quasi-experimental design |
|---|---|---|
| Manipulation of the intervention | Yes | Yes, or a naturally occurring intervention |
| Random assignment | Yes | No |
| Pretest | Optional | Strongly recommended |
| Comparison group | Yes | Usually; some designs use time instead |
| Internal validity | Highest | Moderate; depends on the design |
| Feasibility in real settings | Often low | Often high |
| Typical reporting guideline | CONSORT | TREND |
📱 Ask a statistician about your data — Send a short description of your instrument and sample size on WhatsApp and get a suitable analysis plan. Every analysis comes with the raw SPSS file and a results report.
What are the 6 common quasi-experimental designs?
| Design | Notation | Strength | Weakness |
|---|---|---|---|
| 1. One-group posttest only | X O | Very cheap | Almost no causal evidence; avoid |
| 2. One-group pretest-posttest | O X O | Shows change | Change may come from time, maturation or testing |
| 3. Posttest-only with nonequivalent groups | X O / O | Has a comparison group | Groups may have differed from the start |
| 4. Nonequivalent groups pretest-posttest | O X O / O O | The workhorse for theses | Selection-maturation interaction |
| 5. Interrupted time series | O O O X O O O | Strong when many time points exist | Needs long series; history threats |
| 6. Regression discontinuity | Assignment by a cutoff score | Close to a randomized experiment near the cutoff | Needs a large sample around the cutoff |
In the notation, X is the intervention and O an observation. Designs 1 and 2 are usually called pre-experimental and give weak evidence; design 4 is the most common quasi-experimental design in master’s theses because it is feasible and allows statistical adjustment for pretest differences.
What threats to validity must you address?
Campbell and Stanley listed the classic threats to internal validity, and examiners still ask about them:
- Selection: the groups differed before the intervention.
- History: another event happened at the same time (a new curriculum, a policy change).
- Maturation: participants changed naturally over time.
- Testing: taking the pretest changed the posttest scores.
- Instrumentation: the measure or the observers changed.
- Regression to the mean: groups chosen for extreme scores drift toward the average.
- Attrition: different dropout rates between groups.
For each threat, state in the methodology chapter what you did: a pretest to check baseline equivalence, a comparison group to control history and maturation, the same instrument at both points, and an attrition analysis.
How should you analyze a quasi-experimental design?
The analysis follows the design, not the other way round.
| Design | Typical analysis |
|---|---|
| One-group pretest-posttest | Paired samples t-test or Wilcoxon signed-rank test |
| Nonequivalent groups pretest-posttest | ANCOVA with the pretest as covariate, or mixed (split-plot) ANOVA with time x group interaction |
| Posttest-only nonequivalent groups | Independent samples t-test with covariates, or propensity score methods |
| Interrupted time series | Segmented regression or ARIMA-type models |
| Regression discontinuity | Local regression around the cutoff |
Report effect sizes and confidence intervals, not only p-values. If the pretest shows a large baseline difference, say so and explain how the analysis adjusts for it; never claim the groups were equivalent without testing.
A typical thesis scenario (illustrative)
The following is a constructed example, not a real study. A nursing student wants to test whether a structured discharge-education program reduces readmission anxiety. The hospital introduces the program on one ward only; a similar ward continues usual care. Patients on both wards complete an anxiety scale at admission (pretest) and at discharge (posttest). This is a nonequivalent groups pretest-posttest quasi-experimental design. The student compares baseline characteristics and pretest scores, runs ANCOVA on posttest anxiety with the pretest as covariate, reports partial eta squared with a confidence interval for the adjusted difference, and discusses selection and history (for example, a staffing change on one ward) as limitations.
How do you write the design in a thesis?
A clear methodology section for a quasi-experimental design names the design (“a nonequivalent control group pretest-posttest design”), explains why randomization was not possible, describes how groups were formed, lists the measures and their reliability, states the timing of measurements and the intervention, and addresses each validity threat. The TREND statement (Des Jarlais and colleagues, 2004) gives a 22-item checklist for reporting nonrandomized evaluations, and using it makes the chapter easier to examine.
What claims can you make from the results?
Be modest. A quasi-experimental design supports a statement such as “students in the intervention classes improved more than those in comparison classes, after adjusting for pretest scores”. It rarely supports “the method causes higher achievement in all students”. Discuss the threats you could not rule out in the limitations section, and suggest a randomized design for future research where it would be feasible.
📱 Ask on WhatsApp now — 5 years of experience across all disciplines and 14 academic services. Two-stage payment (80/20); revisions are free until supervisor approval, within 30% of the chapter.
Frequently asked questions
Is a quasi-experimental design quantitative?
Yes, it is a quantitative design because it measures outcomes numerically and compares them statistically. It can be combined with interviews in a mixed-methods thesis to explain why the intervention worked or not.
What is the difference between quasi-experimental and correlational research?
In a quasi-experiment an intervention is introduced or compared deliberately and outcomes are measured before and after. Correlational research measures existing variables without any intervention and cannot speak to effects as directly.
Do I need a control group?
A comparison group is strongly recommended. Without one, as in the one-group pretest-posttest design, history and maturation explain the change as well as the intervention does.
How large should each group be?
Base it on a power analysis for the planned test, for example ANCOVA with the expected effect size. Intact classes often fix the size, so report the power you achieved.
Can I call my study an experiment if I could not randomize?
Call it a quasi-experimental design. Examiners view mislabeling a nonrandomized study as a true experiment as a methodological error.
Sources
- Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Houghton Mifflin.
- Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Rand McNally.
- Harris, A. D., McGregor, J. C., Perencevich, E. N., Furuno, J. P., Zhu, J., Peterson, D. E., & Finkelstein, J. (2006). The use and interpretation of quasi-experimental studies in medical informatics. Journal of the American Medical Informatics Association, 13(1), 16-23. https://doi.org/10.1197/jamia.M1749
- Des Jarlais, D. C., Lyles, C., Crepaz, N., & the TREND Group. (2004). Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: The TREND statement. American Journal of Public Health, 94(3), 361-366. https://doi.org/10.2105/AJPH.94.3.361
- Schulz, K. F., Altman, D. G., & Moher, D. (2010). CONSORT 2010 statement: Updated guidelines for reporting parallel group randomised trials. BMJ, 340, c332. https://doi.org/10.1136/bmj.c332
Sources last checked: November 2026. This article is general educational guidance and does not replace your university regulations or your supervisor’s instructions. How we prepare our content.