How many people do I need for a regression?

It depends on the number of predictors and the effect you expect. Use the regression sample size calculator: with 3 predictors and a medium effect you need 77 people for 0.80 power.

Can I use Likert scale variables?

Yes, if you use each scale’s total or mean score (not a single item). That is standard practice.

What if my predictor is a category, like sex?

Code it as 0 and 1 (dummy variable). With more than two categories, create a 0/1 column for each category but one.

Yes. Coefficients, standard errors, p-values, R², F, VIF and Durbin-Watson match SPSS (Analyze > Regression > Linear, Enter method) and Python (statsmodels).

All tools

Simple and multiple linear regression calculator

Linear regression tells you how much one or more variables (predictors) explain another (the dependent variable), and how much each one weighs. Paste your data from Excel: one row per person, the first column is the variable you want to explain (Y) and the next ones are the predictors. If the first row has the variable names, they are used in the table.

Paste your data (first column = Y, the rest = predictors)

The calculation runs in your browser. We don’t store what you type.

How it’s calculated

The model is estimated by ordinary least squares: Y = B0 + B1·X1 + B2·X2 + … + error.

R² is the share of Y’s variance the predictors explain. Adjusted R² corrects it for the number of predictors.

The F test says whether the model as a whole explains anything. Each coefficient’s t says whether that predictor adds something, controlling for the others.

Beta (β) is the standardized coefficient: it lets you compare which predictor weighs more even on different scales. VIF checks multicollinearity and Durbin-Watson the independence of residuals.

How to read the result

If the F test has p < .05, the model explains a significant part of the dependent variable. See how much with R²: for example, R² = .39 is 39%.
A predictor is significant if its p is below .05. The sign of B gives the direction: positive, more X more Y; negative, more X less Y.
Compare the βs to see which predictor weighs more. A VIF above 5 (or 10, depending on the author) means the predictors are too similar.
Durbin-Watson near 2 means independent residuals; below 1.5 or above 2.5 is worth checking.

Next stepSample size for correlation and regressionIs your sample big enough for this regression? Calculate the minimum size from the number of predictors.

Frequently asked questions

How many people do I need for a regression?: It depends on the number of predictors and the effect you expect. Use the regression sample size calculator: with 3 predictors and a medium effect you need 77 people for 0.80 power.
Can I use Likert scale variables?: Yes, if you use each scale’s total or mean score (not a single item). That is standard practice.
What if my predictor is a category, like sex?: Code it as 0 and 1 (dummy variable). With more than two categories, create a 0/1 column for each category but one.
Does it match SPSS?: Yes. Coefficients, standard errors, p-values, R², F, VIF and Durbin-Watson match SPSS (Analyze > Regression > Linear, Enter method) and Python (statsmodels).

Know what you want to ask?

Build the survey in SurvIQ, share it by link, WhatsApp or QR code, and watch results come in live. The Free plan is free forever.

Create a free survey

More tools

Consistency matrix generatorFill in problems, objectives, hypotheses, variables and methodology, and download the matrix as Word.Operationalization of variables generatorVariables, definitions, dimensions, indicators, items and scale, ready in Word.Sample size calculatorHow many responses you need for results you can trust.Sample size for correlation and regressionHow many people you need for a correlation or multiple regression, with the same calculation as G*Power.Margin of error calculatorHow precise the results from the responses you already have are.Cronbach’s alpha calculatorCheck your questionnaire’s reliability with your pilot test data.