R (Correlation Coefficient)
What it measures: How strongly two things are related, in this case, the relationship between predicted property values from the model and actual sale prices.
How to think about it: If R is close to 1, the model’s predictions closely follow real sale prices.
Example: R = 0.95 means very strong alignment between the model and market sales.
R² (Coefficient of Determination)
What it measures: The percentage of variation in sale prices that the model can explain.
How to think about it: The higher the number (closer to 1.00), the better the model fits the data.
Example: R² = 0.90 means 90% of the differences in property sale prices are explained by the model.
Adjusted R²
What it measures: Similar to R², but it penalizes the model for using too many variables, preventing overfitting.
Why it matters: It gives a more realistic picture of how well the model works, especially when lots of factors are included.
Standard Error of the Estimate (Std. Error)
What it measures: On average, how far off the model’s predicted value is from actual sale prices.
How to think about it: Lower numbers are better because it means the predictions are closer to the real sales.
Mean Ratio
What it measures: The average ratio of assessed value divided by sale price.
How to think about it: 1.00 means the assessed values match the market perfectly on average. Over 1.00 means assessments are higher than sales prices. Under 1.00 means assessments are lower than sales prices.
Median Ratio
What it measures: The middle ratio of all properties, where half are above and half are below.
Why it’s important: It is less affected by unusual, very high, or very low sales.
Weighted Mean Ratio
What it measures: Like the mean ratio, but big sales have more influence than small ones.
Why it’s useful: Helps show if high-value properties are being assessed fairly compared to lower-value properties.
COD (Coefficient of Dispersion)
What it measures: How consistent or uniform assessments are across similar properties.
How to think about it: Lower COD means more uniform and fair. IAAO standards often recommend CODs below 15% for residential areas.
PRD (Price Related Differential)
What it measures: Whether high-value properties are assessed differently than low-value ones.
Ideal number: Close to 1.00. Above 1.03 may mean lower-value homes are assessed too high relative to expensive ones. Below 0.98 may mean higher-value homes are assessed too high.
PRB (Price Related Bias)
What it measures: A more modern, statistical way of seeing if there is bias between low and high-value properties.
Target range: Between -0.05 and +0.05. Positive numbers suggest low-value homes might be assessed higher than high-value homes. Negative numbers suggest the reverse.
COV (Coefficient of Variation)
What it measures: How much ratios vary as a percentage of the average ratio.
Why it’s useful: Lower COV means less variation and better consistency. It is similar to COD but expressed as a percentage of the average.
Relative Standard Error
What it measures: The reliability of the mean ratio estimate.
How to think about it: A smaller value means the average ratio is measured with more certainty, like a margin of error in polling.
N (Sales)
What it measures: The number of verified sales included in the study or model.
Why it matters: More sales mean more reliable statistics. Very small numbers of sales can produce unstable results.
Time Bias (R)
What it measures: Whether there is a trend in sale prices over time that is not accounted for.
Example: If sale prices are increasing rapidly but assessments do not reflect that, there is a positive time bias. If prices are falling but assessments are still high, there is a negative time bias.