A regression of monthly sales ($) on advertising spend ($) gives R2=0.55. Write a one-sentence interpretation.
Example 3
hard
Suppose r=0.30 for one dataset and r=−0.30 for another. Compare their R2 values and the direction of the relationship.
Example 4
challenge
Two studies report R2=0.20 on n=100,000 and R2=0.85 on n=8. Which model is 'better'?
Example 5
medium
Total variation in y is 200; SSR (sum of squared residuals) is 50. Find R2.
Example 6
hard
Why does adding more predictors to a regression never decrease R2?
Example 7
challenge
A small data set has yˉ=5 and observed y values 3,5,7,5. The model predicts y^i=4,5,6,5. Find R2.
Example 8
hard
A regression model has R2=0.85. Interpret this value.
Example 9
hard
If the correlation coefficient is r=−0.9, find R2 and interpret both values.
Practice Problems
Try each one, then open the solution to compare your method.
Example 1
easy
A correlation is r=0.9. Find R2.
Example 2
easy
R2=0.64. What percent of the variation in y is explained by the model?
Example 3
easy
R2 ranges between which two values?
Example 4
easy
R2=0.80. What proportion of variation is unexplained?
Example 5
easy
If a model explains all the variation in y, what is R2?
Example 6
easy
If a model explains none of the variation in y, what is R2?When the model explains none of the variation, R²
Example 7
easy
r=−0.7. Find R2.
Example 8
easy
R2=0.25. Express the explained variation as a percent.
Example 9
medium
Model A has R2=0.81; Model B has R2=0.49 on the same data. Which explains more variance, and by how many percentage points?
Example 10
medium
A regression has R2=0.36. Find the magnitude of the correlation coefficient ∣r∣.
Example 11
medium
A model has R2=0.95. Explain why this alone does not guarantee good predictions on new data.
Example 12
medium
Why is comparing R2 between two models built on completely different datasets misleading?
Example 13
medium
R2=0.49 and the regression slope is positive. Find r including its sign.
Example 14
medium
A model explains 70% of variance in y. The total variance of y is 200. How much variance is explained, in the same units?
Example 15
medium
Adding more predictors to a regression raised R2 from 0.82 to 0.83. Why is this not strong evidence the new predictors help?
Example 16
medium
R2=0.9 is reported but the residual plot shows a strong curved pattern. Should you trust the model? Why?
Example 17
medium
A regression reports R2=0.49. A student says 'the model is 49% accurate.' Why is that interpretation wrong?
Example 18
challenge
A simple regression has R2=0.64 and a negative slope. State r, and the percent of variation left unexplained.
Example 19
challenge
Total variance of y is 50. After regression, the residual (unexplained) variance is 20. Find R2.
Example 20
challenge
Two simple regressions: Model P has r=0.6, Model Q has r=0.8. By what factor does Model Q explain more variance than Model P?
Example 21
easy
A correlation is r=0.6. Find R2.
Example 22
easy
A correlation is r=−0.5. Find R2.
Example 23
medium
R2=0.72. Write a one-sentence interpretation in context of predicting test scores from study hours.
Example 24
medium
Total variation in y is SStot=200. Residual variation is SSres=50. Compute R2.
Example 25
medium
Total variation SStot=400, residual SSres=320. Compute R2.
Example 26
medium
R2=0.64. If correlation r is negative, what is r?
Example 27
medium
A model has R2=0.04. Which of the following best describes the fit: strong, moderate, weak, or no linear fit?
Example 28
medium
Total variation in y is 1000. The model leaves 250 unexplained. What is R2 as a percent?
Example 29
hard
A regression has r=0.7 but a clear curved residual plot. Why might using R2=0.49 be misleading?
Example 30
hard
R2=0.90 for predicting house price from square footage. Is it valid to say square footage causes price differences?
Example 31
hard
A linear model has R2=0.36. By how many percentage points does adding a predictor (giving R2=0.45) increase the explained variation?
Example 32
hard
A regression line is y^=2+0.5x and R2=0.64. If we instead predicted y using only the mean yˉ, the prediction errors would be larger by what factor in sum-of-squares?
Example 33
medium
R2 went from 0.25 to 0.81 after fitting a curve instead of a line. Did the new model improve the explained variation? By how much (in percentage points)?
Example 34
medium
If R2=0.16, and the residual sum of squares is 84, what is the total sum of squares?
Example 35
hard
A linear regression of y on x gives R2=0.49. If we instead regress x on y, what is R2?
Example 36
easy
A correlation is r=0.8. Find R2.
Example 37
easy
R2=0.49. What percent of variation in y is explained?
Example 38
easy
R2=0.36. What proportion of variation is unexplained?
Example 39
easy
r=0. Find R2.
Example 40
easy
r=−0.5. Find R2.
Example 41
medium
R2=0.81. Find ∣r∣.
Example 42
medium
Model A: R2=0.72. Model B: R2=0.48. By how many percentage points does A explain more variance?
Example 43
medium
r=0.6 on a sample of n=30 pairs. Find R2.
Example 44
medium
R2=0.04. Find ∣r∣.
Example 45
hard
Total sum of squares is 400; SSR =100. Find R2.
Example 46
hard
A regression has R2=0.999 on the training data but predicts poorly on new data. What problem is most likely?
Example 47
hard
Two regression models are fit on different data sets. Model 1 has R2=0.9; Model 2 has R2=0.7. Can we conclude Model 1 fits its data more accurately?
Example 48
medium
∣r∣=0.7. Find R2.
Example 49
hard
SST =500; SSR =75. Find the percent of variation explained.
Example 50
hard
Two models are compared: Model A has R2=0.72 and Model B has R2=0.58. Which model provides a better fit and why?
Example 51
hard
A linear model has R2=0.64. What percentage of the variation is not explained by the model?
What to focus on
Core idea:Compare the squared misses from the model with the squared misses from guessing the mean: the share removed is R2, a measure of fit on this data that says nothing about direction, about whether the shape is right, or about cause.
Common stuck point:Students often know a procedure related to r-squared (coefficient of determination) but skip the recognition step: Do you need the share of the variation in y that a fitted model accounts for, from 0 to 1? That leads to a calculation or graph that looks reasonable but answers a different question.
Ask yourself:Do you need the share of the variation in y that a fitted model accounts for, from 0 to 1?
Concept recap
R-squared (the coefficient of determination) is the proportion of variance in the dependent variable that is explained by the independent variable(s) in a regression model. It ranges from 0 to 1, where 0 means the model explains none of the variability and 1 means it explains all of it.
R-squared rates a model the way noise-cancelling headphones are rated: the spread of y around its mean is the noise, and R2=0.85 means the model removes 85% of the noise power, the squared spread. A high rating says how much was removed, not whether the model is the true explanation.