An appraiser develops a multiple regression model to estimate residential sale price using GLA, age, number of bathrooms, and distance to nearest highway. The model yields an adjusted R² of 0.72 and a standard error of the estimate of $24,500. Which statement best reflects the appropriate interpretation of the standard error of the estimate in this context?
Correct Answer
B) Approximately 68% of the predicted sale prices fall within ±$24,500 of the actual sale prices, assuming residuals are normally distributed.
Why this is correct: The standard error of the estimate is the standard deviation of the residuals, the vertical distances between the actual sale prices and the prices the model predicts. If the residuals are approximately normally distributed and their spread is stable across the range of predictions, roughly 68% of observations fall within one standard error of the regression surface, so about 68% of predictions land within $24,500 of the actual price. That is the empirical rule applied to residuals, and it is what makes the standard error a usable statement of typical prediction accuracy. Why the other choices are wrong: 'The average difference between observed sale prices and predicted sale prices is $24,500' describes the mean absolute error, a different statistic; a standard deviation is a root-mean-square measure and is larger than the mean absolute deviation for the same data. 'The model explains 72% of the variation in sale price, and the remaining unexplained variation has a standard deviation of $24,500' welds two statistics together incorrectly; adjusted R² is a proportion of variance explained and does not itself quantify residual spread. 'Each coefficient's standard error is $24,500, indicating low precision in estimating the slope parameters' confuses the standard error of the estimate with the standard errors of the individual coefficients, which are separate figures reported per variable and expressed in that variable's units. Exam tip: Keep three regression outputs distinct: R² measures how much variation is explained, the standard error of the estimate measures how far predictions typically miss in dollars, and a coefficient's standard error measures how precisely that one slope is known. An appraiser who uses a model must be competent to read all three.
Why This Is the Correct Answer
Roughly 68 percent of predictions falling within plus or minus one standard error of actual prices, assuming normally distributed residuals, is the correct interpretation and correctly states the assumption it depends on. The standard error is a standard deviation of residuals, so the empirical rule governs the coverage. Stating the normality condition matters because the 68 percent figure does not hold if residuals are skewed or heteroskedastic. For an appraiser the practical implication is that a model with this standard error supports a range far better than a point.
Why the Other Options Are Wrong
Option A: The average difference between observed sale prices and predicted sale prices is $24,500.
Average difference describes the mean absolute error, which is a different and generally smaller statistic than a standard deviation of residuals. Squaring the residuals before averaging gives larger deviations disproportionate influence, so the standard error exceeds the mean absolute error for any realistic distribution. The distinction matters because quoting the standard error as an average error overstates typical accuracy.
Option C: The model explains 72% of the variation in sale price, and the remaining unexplained variation has a standard deviation of $24,500.
The first half of this option correctly describes adjusted R-squared, but the second half improperly ties the residual standard deviation to it. The two statistics are computed differently and neither is derived from the other in the way stated. Pairing a true statement with a false linkage is what makes this the most sophisticated distractor.
Option D: Each coefficient’s standard error is $24,500, indicating low precision in estimating the slope parameters.
Standard errors of individual coefficients measure the precision of each estimated slope and are reported separately, one per variable, in units of that variable. The standard error of the estimate is a single figure describing prediction error in dollars. Confusing the two conflates uncertainty about the model's parameters with uncertainty about its predictions.
SEE Speaks in Dollars
The standard error of the estimate is the only regression statistic that answers how far off in dollars. R-squared is unitless and describes explained variation. Coefficient standard errors describe uncertainty about slopes. Follow the units to the right statistic.
How to use: When a stem quotes a figure in dollars alongside an R-squared, the dollar figure is prediction error and the ratio is explained variation. Match each interpretation to its own statistic and reject options that link them.
Exam Tip
The 68 and 95 percent coverage figures depend on normally distributed residuals. An answer that states the coverage without stating the assumption is weaker than one that names it.
Common Mistakes to Avoid
- -Reading the standard error of the estimate as a mean absolute error
- -Confusing the standard error of the estimate with coefficient standard errors
- -Quoting coverage percentages without checking whether residuals are approximately normal
Concept Deep Dive
Analysis
The standard error of the estimate is the standard deviation of the residuals, meaning the typical distance between what the model predicted and what a property actually sold for. It is expressed in the units of the dependent variable, dollars here, which makes it far more useful to an appraiser than a unitless fit statistic: a standard error of $24,500 says something concrete about how far off a prediction is likely to be. Under the usual assumptions that residuals are normally distributed and of constant variance, the empirical rule applies, so roughly 68 percent of observations fall within one standard error of the regression line and roughly 95 percent within two. On this model that means about two thirds of predictions land within $24,500 of the actual price and about ninety-five percent within $49,000, which for a typical suburban house is a wide band and should temper any claim of precision. The adjusted R-squared of 0.72 answers a different question, describing the share of variance explained after penalizing for the number of variables, and it says nothing directly about the dollar size of a typical error.
Background Knowledge
You need the components of regression output, coefficients and their standard errors, t-statistics, R-squared and adjusted R-squared, and the standard error of the estimate, and what each measures. You should also know the empirical rule for normally distributed data and the assumptions of normality and homoscedasticity underlying inference.
Real-World Application
An appraiser with a model showing a $24,500 standard error on typical $400,000 homes concludes the tool is useful as a corroborating check but not as a primary indication, reports it that way, and relies on a verified sales grid for her conclusion while disclosing the model's performance statistics.
More Statistics Questions
A set of comparable sales has a mean of $250,000 and a standard deviation of $20,000. What is the coefficient of variation?
A property sold for $400,000 and resold three years later for $463,050 with no physical change. What compound annual rate does this indicate?
A histogram of neighborhood sale prices shows two distinct peaks. What does this most likely mean?
What does it mean to validate a regression model?
In a market study, what does a frequency distribution of sale prices show?
An appraiser includes months elapsed since each sale as a variable in a price model. What is this intended to capture?
An appraiser presents a statistical analysis in a report. What must accompany it for the reader to weigh it?
An R-squared of 0.86 in a sales model indicates that:
Which measure would best summarize the most common lot size in a subdivision?
Paired sales analysis and regression differ mainly in that regression:
People Also Study
Real Estate Market
13.6% of exam
Property Description
11.8% of exam
Land or Site Valuation
4.5% of exam
Sales Comparison Approach
16.4% of exam
Cost Approach
13.6% of exam