A linear model separates a constant rate of change from a baseline. Interpreting those two pieces requires knowing the variables, their units, and where the input's zero point is defined.
Slope and intercept answer different questions
For tells the change in per one-unit increase in . Its units are output units divided by input units. The intercept is the predicted output at and has the same units as .
In where is items produced and C is dollars, 18 represents dollars per item and 140 represents a fixed cost. Neither coefficient should be interpreted as a count of items.
Example: recover a model from two observations
A machine's operating cost is 410 dollars for 15 hours and 650 dollars for 27 hours. Assume cost is linear in operating time. The rate is dollars per hour.
Write Using 15 hours gives so The model is At 30 hours it predicts 710 dollars.
The intercept may represent a setup charge if that interpretation fits the pricing arrangement. Two observations alone establish the mathematical intercept under the linear assumption; they do not prove the real business actually charges a setup fee outside the observed range.
Example: a shifted time origin
A population model is where counts years after 2025. The intercept 2,400 refers to 2,025, not calendar year zero. The slope predicts an increase of 80 people per year.
For 2,030, use giving 2,800 people. Substituting 2,030 directly for would ignore the variable definition. If a new variable counts years after 2028, then so the model becomes Changing the time origin changes the intercept but not the rate.
Population is 2400 at year 2025, 2640 at year 2028, and 2800 at year 2030 under the linear model.
The same rate adds 80 people each year. The point at becomes the new intercept when the clock starts in 2028.
Describe this figure
Population is 2,400 at year 2025, 2,640 at year 2028, and 2,800 at year 2030 under the linear model.
Marked points: 2,025: filled marker at (0, 2,400); 2,028: filled marker at (3, 2,640); 2,030: filled marker at (5, 2,800).
Proportional is stronger than linear
A proportional model has form so its graph passes through the origin and its nonzero input-output ratios are constant. A general linear model with nonzero intercept does not have that property.
For doubling does not generally double . At equal to 4, is 24; at equal to 8, is 36. The same 12-unit baseline is present in both outputs. Constant differences per input unit indicate linearity; constant ratios of output to input indicate proportionality.
Interpret signs and feasible inputs
For the negative slope indicates loss of 25 units per time unit. The model reaches zero at If W is water remaining, the reasonable domain is the negative continuation is not physically meaningful.
When an input counts items or people, restrict it to integers. A continuous line can be a useful algebraic model even when only isolated points correspond to feasible choices. Also distinguish prediction within observed values from extrapolation beyond them, where the constant-rate assumption may fail.
Try it yourself
A linear model gives 62 dollars at 4 units and 98 dollars at 10 units. Find the equation and interpret its coefficients. Is it proportional?
Answer: The slope is dollars per unit. Substitution gives so The model is with a 38-dollar zero-input value. It is not proportional because that intercept is nonzero.
What to remember
Define the input's units and zero point before interpreting coefficients. Separate baseline from rate, distinguish proportionality from linearity, and state the domain where the model makes sense.