Statistical methods · Dissertation topic selection
50 practical dissertation questions grouped by regression model, each with an outcome, predictors, accessible data and a related university thesis.
50 Research questions
10 Model groups
17 Dataset sources
4 Universities cited
2020–26 Thesis years
By Siddharth Gupta · Statssy · Updated October 2026
Most students pick a topic first and worry about the model later. Then they discover that their outcome is a count or a ranked scale, and the original plan needs revision.
Start by defining your outcome, selecting the model suited to it and confirming the dataset contains the variables you need. The 50 examples below help you move from a regression model to a feasible research question, with related theses to consult.
Which regression do you need? Start with your outcome and study design. These are starting points; model assumptions and data structure determine the final choice.
How to use this list Identify whether your outcome is continuous, binary, ordinal, nominal or a count. Choose two or three topics and open the associated dataset. Confirm every proposed predictor is available and has the required measurement level. Read the linked thesis for examples of modelling and reporting. Book a free discussion if you need to compare models.Jump to a regression model 1. Simple linear regression 2. Multiple linear regression 3. Binary logistic regression 4. Ordinal and multinomial logistic regression 5. Poisson and negative binomial regression 6. Panel data regression (fixed and random effects) 7. Multilevel (mixed effects) regression 8. Time series regression 9. Interaction and moderation models 10. Quantile regression Regression model 01 / 10
Simple linear regression Use it when you have one continuous outcome and one predictor. It is the right starting point for undergraduate projects and the first model in most Masters chapters.
TOPIC 01 Does age predict yearly medical charges? Field Health economics Outcome Medical charges Predictors Age Dataset Medical Cost Personal (Kaggle) Free login TOPIC 02 How much does each extra square foot add to a house’s sale price? Field Real estate Outcome Sale price Predictors Living area Dataset House Prices (Kaggle) Free login TOPIC 03 Does alcohol content predict a wine’s quality score? Field Food science Outcome Quality score Predictors Alcohol content Dataset Wine Quality (UCI) Open TOPIC 04 Does GDP per head predict life expectancy across countries? Field Development economics Outcome Life expectancy Predictors GDP per head (log) Dataset World Bank WDI Open TOPIC 05 Do first-term grades predict final grades? Field Education Outcome Final grade Predictors First-term grade Dataset Student Performance (UCI) Open Regression model 02 / 10
Multiple linear regression Use it when a continuous outcome has several predictors. It lets you show the effect of one factor while holding the others constant.
TOPIC 06 Do smoking, BMI and age together explain medical charges? Field Health economics Outcome Medical charges Predictors Smoking, BMI, age, children, region Dataset Medical Cost Personal (Kaggle) Free login TOPIC 07 Which house features add the most to sale price? Field Real estate Outcome Sale price (log) Predictors Area, quality, age, garage, neighbourhood Dataset House Prices (Kaggle) Free login TOPIC 08 Which chemical properties predict wine quality? Field Food science Outcome Quality score Predictors Acidity, sugar, sulphates, alcohol Dataset Wine Quality (UCI) Open TOPIC 09 Do study time, absences and family support predict final grades? Field Education Outcome Final grade Predictors Study time, absences, family support, failures Dataset Student Performance (UCI) Open TOPIC 10 Does education still raise wages once experience is accounted for? Field Labour economics Outcome Hourly wage (log) Predictors Years of schooling, experience, sex, region Dataset PSID Free login Regression model 03 / 10
Binary logistic regression Use it when the outcome is yes or no. It is the most common model in health and business dissertations.
TOPIC 11 Which risk factors predict heart disease within ten years? Field Cardiology Outcome Heart disease in 10 years (yes or no) Predictors Age, smoking, cholesterol, blood pressure, diabetes Dataset Framingham Heart Study (Kaggle) Free login TOPIC 12 Does being born abroad change the odds of earning above 50K? Field Labour economics Outcome Income above 50K (yes or no) Predictors Native country, education, hours, sex, age Dataset Adult Income (UCI) Open TOPIC 13 Which bank clients say yes to a term deposit? Field Marketing Outcome Subscribed (yes or no) Predictors Age, job, balance, contact type, previous outcome Dataset Bank Marketing (UCI) Open TOPIC 14 Does chest pain type predict heart disease after adjusting for age and sex? Field Cardiology Outcome Heart disease (yes or no) Predictors Chest pain type, age, sex, resting blood pressure Dataset Heart Disease (UCI) Open TOPIC 15 At what cholesterol level does heart disease risk reach 50 percent? Field Cardiology Outcome Heart disease (yes or no) Predictors Serum cholesterol Dataset Heart Disease (UCI) Open Regression model 04 / 10
Ordinal and multinomial logistic regression Use ordinal regression when the outcome is ranked, like a 1 to 5 scale. Use multinomial regression when the outcome has unordered categories, like a choice between options.
TOPIC 16 Does trust in parliament depend on age and education? Field Political science Outcome Trust level (ordered) Predictors Age, education, income, country Dataset European Social Survey Free login TOPIC 17 Do unemployed people rate their own health worse? Field Labour and health Outcome Self-rated health (ordered) Predictors Employment status, age, sex, income Dataset European Social Survey Free login TOPIC 18 Does alcohol content still predict wine quality when quality is treated as ranked groups? Field Food science Outcome Quality band (low, medium, high) Predictors Alcohol, acidity, sulphates Dataset Wine Quality (UCI) Open TOPIC 19 Which factors predict the party a person votes for? Field Political science Outcome Party voted for (categories) Predictors Age, education, income, religion, region Dataset European Social Survey Free login TOPIC 20 Does time spent with friends predict weekend alcohol use levels in students? Field Education and health Outcome Weekend alcohol use (1 to 5) Predictors Going out, study time, family relations, sex Dataset Student Performance (UCI) Open Regression model 05 / 10
Poisson and negative binomial regression Use these when the outcome is a count, like visits, absences or cases. Switch to negative binomial when the variance is much bigger than the mean.
TOPIC 21 Does weather predict daily bike rentals? Field Transport Outcome Daily rentals (count) Predictors Temperature, humidity, wind, season Dataset Bike Sharing (UCI) Open TOPIC 22 Do family relationships predict how many classes a student misses? Field Education Outcome School absences (count) Predictors Family relations, study time, alcohol use, sex Dataset Student Performance (UCI) Open TOPIC 23 Do crime counts rise with unemployment across Indian states? Field Criminology Outcome Reported crimes (count) Predictors Unemployment rate, literacy, urbanisation Dataset Open Government Data India Open TOPIC 24 Do bike rentals fall more on holidays than on weekends? Field Transport Outcome Daily rentals (count) Predictors Holiday, weekday, weather Dataset Bike Sharing (UCI) Open TOPIC 25 Do countries with more doctors per head have fewer maternal deaths? Field Global health Outcome Maternal deaths (count, with births as offset) Predictors Doctors per 1,000, health spending, GDP per head Dataset World Bank WDI Open View the printable 50-topic guide
Regression model 06 / 10
Panel data regression (fixed and random effects) Use panel models when you follow the same countries, firms or people over several years. Fixed effects remove anything about a unit that does not change over time.
TOPIC 26 Does foreign direct investment raise growth once country effects are removed? Field Development economics Outcome GDP growth Predictors FDI inflows, trade, investment, inflation Dataset World Bank WDI Open TOPIC 27 Does capital per worker explain productivity differences across countries? Field Macroeconomics Outcome Output per worker (log) Predictors Capital per worker, human capital index Dataset Penn World Table Open TOPIC 28 Does lower corruption make health spending more effective? Field Public policy Outcome Life expectancy Predictors Health spending, corruption index, interaction Dataset QoG Data Open TOPIC 29 Does losing a job lower household income years later? Field Labour economics Outcome Household income (log) Predictors Job loss, years since loss, family size Dataset PSID Free login TOPIC 30 Does air pollution lower labour productivity across countries? Field Environmental economics Outcome GDP per worker (log) Predictors PM2.5 exposure, capital, education Dataset World Bank WDI Open Regression model 07 / 10
Multilevel (mixed effects) regression Use multilevel models when people sit inside groups, like students in schools or people in countries. They split the variation between the two levels.
TOPIC 31 How much of the gap in maths scores lies between schools? Field Education Outcome Maths score Predictors Student background, school type, school resources Dataset PISA Data (OECD) Open TOPIC 32 Do country inequality and personal income both shape happiness? Field Wellbeing economics Outcome Happiness score Predictors Personal income, country Gini index Dataset European Social Survey Free login TOPIC 33 Does a school’s social mix predict reading scores beyond a student’s own background? Field Education Outcome Reading score Predictors Student SES, school average SES Dataset PISA Data (OECD) Open TOPIC 34 Do district factors explain state differences in literacy across India? Field Development Outcome District literacy rate Predictors Urbanisation, school access, state Dataset Open Government Data India Open TOPIC 35 Does unemployment hurt wellbeing more in countries with weaker welfare systems? Field Wellbeing economics Outcome Life satisfaction Predictors Unemployment, country social spending, cross-level interaction Dataset European Social Survey Free login Regression model 08 / 10
Time series regression Use time series models when one unit is measured many times, like monthly inflation. Check for trends and autocorrelation before you trust any result.
TOPIC 36 Do oil price shocks raise US inflation? Field Macroeconomics Outcome CPI inflation Predictors Oil price, lagged inflation, interest rate Dataset FRED Open TOPIC 37 How does US unemployment respond to interest rate rises? Field Macroeconomics Outcome Unemployment rate Predictors Federal funds rate and its lags Dataset FRED Open TOPIC 38 Did COVID-19 break the trend in US unemployment? Field Labour economics Outcome Monthly unemployment rate Predictors Time, pandemic indicator, time after pandemic Dataset FRED Open TOPIC 39 Did India’s growth trend change after a major policy shift? Field Development economics Outcome Annual GDP growth Predictors Time, policy indicator, time after policy Dataset World Bank WDI Open TOPIC 40 Can weather data forecast hourly bike demand? Field Transport Outcome Hourly rentals Predictors Lagged rentals, temperature, hour, season Dataset Bike Sharing (UCI) Open Regression model 09 / 10
Interaction and moderation models Use an interaction term when the effect of one variable depends on another. It turns a simple question into a much stronger dissertation.
TOPIC 41 Does smoking raise medical charges more for people with a high BMI? Field Health economics Outcome Medical charges Predictors Smoking, BMI, smoking by BMI Dataset Medical Cost Personal (Kaggle) Free login TOPIC 42 Is the return to education different for men and women? Field Labour economics Outcome Income above 50K (yes or no) Predictors Education, sex, education by sex Dataset Adult Income (UCI) Open TOPIC 43 Does house quality matter more in some neighbourhoods than others? Field Real estate Outcome Sale price (log) Predictors Overall quality, neighbourhood, quality by neighbourhood Dataset House Prices (Kaggle) Free login TOPIC 44 Does the link between family background and test scores differ by gender? Field Education Outcome Maths score Predictors Family SES, gender, SES by gender Dataset PISA Data (OECD) Open TOPIC 45 Does diabetes raise heart disease risk more in smokers? Field Cardiology Outcome Heart disease in 10 years (yes or no) Predictors Diabetes, smoking, diabetes by smoking Dataset Framingham Heart Study (Kaggle) Free login Regression model 10 / 10
Quantile regression Use quantile regression when you care about the top or bottom of a distribution, not only the average. It suits income, cost and price questions.
TOPIC 46 Does education raise wages more at the top of the wage ladder? Field Labour economics Outcome Hourly wage (log) Predictors Years of schooling, experience, sex Dataset PSID Free login TOPIC 47 Do smoking costs rise most among the highest spenders? Field Health economics Outcome Medical charges Predictors Smoking, age, BMI Dataset Medical Cost Personal (Kaggle) Free login TOPIC 48 Does growth help the poorest countries as much as the richest? Field Development economics Outcome Income share of bottom 20 percent Predictors GDP growth, inflation, trade Dataset World Bank WDI Open TOPIC 49 Do bigger homes add more value at the top of the market? Field Real estate Outcome Sale price Predictors Living area, quality, age Dataset House Prices (Kaggle) Free login TOPIC 50 Is the gender earnings gap wider at the top or at the bottom? Field Labour economics Outcome Hourly wage (log) Predictors Sex, education, experience Dataset PSID Free login View the printable 50-topic guide
Printable regression topic guide
Get all 50 topics in one PDF The printable guide will contain all 50 questions, models, outcomes, predictors, dataset links and thesis references. The download link will appear here after the PDF is created and uploaded.
PDF download coming soon How to turn one of these into your own topic Describe your outcome. Identify its measurement scale and the sampling design.Validate the variables. Confirm outcome and predictors in the codebook.Plan sample size. Use a power or precision calculation rather than a fixed cases-per-predictor rule.Check model assumptions. For linear models assess functional form, residual diagnostics and collinearity; logistic models also require attention to collinearity, sparse data and separation.Make the analysis purposeful. Add interactions, hierarchical structure or quantiles only when the research question supports them.For more guidance, see choosing dissertation statistics , binary versus multinomial logistic regression , and ordinal versus linear regression for Likert responses .
More research topic collections Need a second pair of eyes?
Pick a model that fits your data. Bring your outcome variable and topic shortlist for a free 15-minute discussion of regression methods and data suitability.
Book a free 15-minute call 