What are the conventional f2 values for small, medium, and large effect sizes in regression?

What are the conventional f2 values for small, medium, and large effect sizes in regression?

• A. 0.02, 0.15, 0.35
• B. 0.01, 0.05, 0.10
• C. 0.10, 0.20, 0.30
• D. 0.05, 0.10, 0.20

Answer: (A)

In regression, f squared measures how much a set of predictors changes the amount of variance explained by the model relative to what remains unexplained. It’s calculated as f^2 = ΔR^2 / (1 − R^2_full), and it helps categorize the strength of that predictor set’s impact on the outcome. Cohen provided conventional benchmarks to guide interpretation: small around 0.02, medium around 0.15, and large around 0.35. These thresholds give a sense of whether adding a group of predictors yields only a tiny improvement, a noticeable/moderate improvement, or a substantial one in model fit. The values 0.02, 0.15, and 0.35 match these standard cutoffs, making them the best representation of conventional f^2 sizes for small, medium, and large effects. The other sets do not align with these widely used benchmarks.

In regression, f squared measures how much a set of predictors changes the amount of variance explained by the model relative to what remains unexplained. It’s calculated as f^2 = ΔR^2 / (1 − R^2_full), and it helps categorize the strength of that predictor set’s impact on the outcome. Cohen provided conventional benchmarks to guide interpretation: small around 0.02, medium around 0.15, and large around 0.35. These thresholds give a sense of whether adding a group of predictors yields only a tiny improvement, a noticeable/moderate improvement, or a substantial one in model fit. The values 0.02, 0.15, and 0.35 match these standard cutoffs, making them the best representation of conventional f^2 sizes for small, medium, and large effects. The other sets do not align with these widely used benchmarks.