How to report regression results in a dissertation

Problem

Students often know how to run a regression model but struggle to report it properly. A dissertation should not simply paste software output into the results chapter. Regression reporting requires a clear explanation of the model purpose, the outcome, predictors, adjustment variables, estimates, confidence intervals, p-values, assumptions and limitations. Without this structure, readers cannot understand what the model answers or how reliable the interpretation is.

• Regression tables are copied directly from software.
• The outcome and predictors are not clearly described.
• Adjusted and unadjusted results are mixed without explanation.
• Coefficients are reported without interpretation.
• Confidence intervals are omitted.
• P-values are treated as the main result.
• Assumption checks are not mentioned.
• Causal language is used without justification.

Intuition

A regression model is a formal way of describing how an outcome is associated with one or more predictors. Reporting regression is therefore not only about numbers. It is about explaining the question, the model, the estimate and the uncertainty. The reader should understand what was modelled, why it was modelled, what the key coefficient means and how cautious they should be when interpreting it.

• The outcome tells the reader what the model is trying to explain.
• The main predictor tells the reader what association is being studied.
• Covariates explain what has been adjusted for.
• The coefficient or odds ratio describes the estimated association.
• The confidence interval shows uncertainty around the estimate.
• The p-value should support interpretation, not replace it.

Method

A good regression report follows a predictable structure. First, describe the purpose of the model. Second, define the outcome and predictors. Third, state whether the model is unadjusted or adjusted. Fourth, present estimates with confidence intervals. Fifth, interpret the key findings in plain language. Sixth, mention assumption checks and limitations. This makes the analysis transparent and defensible.

• Step 1: State the aim of the regression model.
• Step 2: Identify the outcome variable and its scale.
• Step 3: Identify the main predictor or exposure.
• Step 4: List adjustment variables and justify them briefly.
• Step 5: Report the model type, such as linear or logistic regression.
• Step 6: Present estimates with 95% confidence intervals.
• Step 7: Interpret the main estimate in the units of the outcome.
• Step 8: Report p-values without making them the only focus.
• Step 9: Mention diagnostics, assumptions or model checks.
• Step 10: Discuss limitations such as confounding, sample size and missing data.

Working

For linear regression, a coefficient usually represents the expected difference or change in the outcome for a one-unit increase in the predictor, holding other variables constant in an adjusted model. For logistic regression, an exponentiated coefficient is an odds ratio, which describes the association with the odds of the event. The interpretation must always match the model type and variable coding.

• Linear regression example: a coefficient of 2.4 for study hours means the expected exam score is 2.4 points higher per additional study hour, if the model is correctly specified.
• Adjusted linear regression example: the coefficient is interpreted while holding included covariates constant.
• Logistic regression example: an odds ratio of 1.50 means higher odds of the event, not necessarily 50% higher probability.
• For categorical predictors, coefficients are interpreted compared with a reference category.
• For binary predictors, the estimate compares the coded group with the reference group.
• For continuous predictors, the unit of measurement must be stated.
• For transformed variables, interpretation should explain the transformed scale carefully.

Limitations

Regression reporting can look precise even when the underlying analysis has important weaknesses. A polished table does not fix poor design, confounding, small sample size, measurement error or missing data. Dissertation writing should clearly separate what the model estimates from what the study design can support.

• Regression cannot remove bias from poor measurement.
• Adjustment only helps for variables included and measured appropriately.
• Too many predictors can overfit small datasets.
• Collinearity can make individual coefficients unstable.
• Model assumptions affect standard errors and inference.
• Observational regression should not be reported as proof of causation.
• Odds ratios can be difficult to interpret when outcomes are common.

Discussion

The discussion should translate regression output into academic meaning. Students should explain whether the findings support the research question, whether the effect size is meaningful, whether the confidence interval is precise and how limitations affect interpretation. The strongest dissertations show statistical understanding rather than simply listing significant predictors.

• Interpret the most important estimate in words.
• Discuss the direction, magnitude and uncertainty of the association.
• Compare findings with existing literature where appropriate.
• Explain whether adjustment changed the estimate meaningfully.
• Avoid focusing only on significant variables.
• Discuss residual confounding and study design limitations.
• State whether the model was explanatory, predictive or descriptive.

How this connects to learning

Use the guide as a bridge between theory and application.

A resource guide should not replace a full course or live teaching session. Instead, it helps you organise your thinking. Use it to identify what you understand, what feels unclear, and what questions you should ask before applying a method to real data.