Sample size for linear regression (subjects per variable)

A quick feasibility floor for a multiple regression: enough participants to support the number of predictors you plan to adjust for.

Method & formula

For a continuous outcome, this calculator applies Green's (1991) rules of thumb: N ≥ 50 + 8k to test the overall model and N ≥ 104 + k to test individual predictors, taking whichever is larger for k predictors. With five predictors that is 109 participants.

A rule of thumb, not a power calculation. These floors assume medium effect sizes. If you can state the effect you care about — an expected R², or a standardized coefficient for one exposure — a proper power analysis (f² approach, Cohen 1988) is more defensible and sometimes demands substantially more. Use this to rule out infeasible plans, not to prove feasible ones.

Source: Green SB. How many subjects does it take to do a regression analysis? Multivariate Behav Res 1991;26:499–510 · Cohen J. Statistical Power Analysis for the Behavioral Sciences, 2nd ed. 1988.

Inputs, explained

Number of predictors/covariates: All model terms: the exposure of interest plus every adjustment variable, counting k−1 terms for a k-category variable.

A sample size is one line of a statistical plan. MedStatica turns a plain-language description of your study into the full plan — design, recommended analysis, power, causal diagram, methods paragraph, and runnable R, Python, SAS, and Stata code. The first plans are free, no card required.

Plan my study free →