Computer regression output worksheet with answer key

8 problems on computer regression output, ready to print for class or homework. The answer key prints on its own page, so hand out everything before it.

Problem 1

A ceramics studio logs 24 kiln firings, recording how long each firing ran, in hours, and what it added to the electricity bill, in dollars. The firings ran from 4 to 12 hours. The software returns:

PredictorCoefSE CoefTP
Constant12.507.55201.660.112
Hours6.400.91696.980.000

S = 8.795, R-Sq = 68.9%, R-Sq(adj) = 67.5%

(a) Write the least-squares regression line, defining your variables. (b) Say exactly which row and column each coefficient came from. (c) The owner reads the Constant row as the electricity cost of a firing that lasts no time at all. Explain the problem with that reading.

Problem 2

A dog daycare tracks 27 business days, recording the number of dogs boarded and the water used that day, in gallons. The output is:

PredictorCoefSE CoefTP
Constant41.0013.70352.990.006
Dogs3.750.60176.230.000

S = 18.41, R-Sq = 60.8%, R-Sq(adj) = 59.2%

(a) Interpret the slope in context, using every part of the standard template. (b) A staff member shortens it to "each extra dog uses 3.75 gallons". Name what that leaves out and why it matters. (c) Two more dogs are booked for tomorrow. By how much does the model predict the day's water use will rise?

Problem 3

A vineyard measures sugar content, in degrees Brix, on 22 sampling dates and regresses it on the number of days since veraison (the point when the grapes start to color). The sampled dates ran from 10 to 40 days since veraison. The output is:

PredictorCoefSE CoefTP
Constant15.500.6986222.190.000
Days0.280.0266710.500.000

S = 0.9778, R-Sq = 84.6%, R-Sq(adj) = 83.8%

(a) Predict the sugar content at 30 days since veraison. (b) Is that prediction interpolation or extrapolation? (c) A grower asks for the predicted sugar content at 60 days. Compute what the line says and explain why you would not report it.

Problem 4

A commercial laundry regresses drying time, in minutes, on load weight, in pounds, for 26 loads weighing between 12 and 48 pounds. The output is:

PredictorCoefSE CoefTP
Constant18.005.68133.170.004
Weight1.500.18178.260.000

S = 8.176, R-Sq = 74.0%, R-Sq(adj) = 72.9%

One 40-pound load actually dried in 70 minutes. (a) Find that load's residual. (b) Say what the sign of the residual means here. (c) Is a miss of this size unusual for this model? Justify your answer with a number from the output.

Problem 5

A transit agency regresses a route's average bus speed, in miles per hour, on the number of stops per mile along that route, for 32 routes with between 2 and 7 stops per mile. The output is:

PredictorCoefSE CoefTP
Constant21.001.812611.590.000
Stops−2.40-2.400.4346−5.52-5.520.000

S = 2.904, R-Sq = 50.4%, R-Sq(adj) = 48.7%

(a) Find the correlation rr between stops per mile and average speed. (b) Explain how you settled the sign, and why R-Sq on its own could never settle it. (c) An analyst says the T value of −5.52-5.52 would have told them the sign just as well. Are they right?

Problem 6

A university regresses a dormitory's monthly water use, in thousands of gallons, on the number of residents, for its 18 dorm buildings. The output is:

PredictorCoefSE CoefTP
Constant14.009.619881.460.165
Residents0.850.0771311.020.000

S = 11.13, R-Sq = 88.4%, R-Sq(adj) = 87.7%

(a) Interpret R-Sq in context. (b) What does the rest of the percentage represent? (c) Two students misread the number. One writes that 88.4 percent of the buildings fall on the line. The other writes that 88.4 percent of a building's water use is explained by its residents. Correct each one.

Problem 7

A regional airline regresses taxi-out time, in minutes, on the number of aircraft ahead in the departure queue, for 40 departures with between 1 and 12 aircraft ahead. The output is:

PredictorCoefSE CoefTP
Constant9.501.98014.800.000
Queue2.200.30527.210.000

S = 4.765, R-Sq = 57.8%, R-Sq(adj) = 56.7%

(a) Which number in the output is the standard deviation of the residuals, and what are its units? (b) Interpret it in context. (c) A student points at 0.3052 and calls that the typical prediction error. Explain what 0.3052 actually measures. (d) Predict the taxi-out time for a departure with 8 aircraft ahead, and attach the typical error to your prediction.

Problem 8

A coastal outfitter regresses the number of kayaks rented in a day on the midday wind speed, in miles per hour, across 35 summer days with wind speeds from 3 to 22 mph. The output is:

PredictorCoefSE CoefTP
Constant82.003.711622.090.000
Wind−3.50-3.500.3453−10.14-10.140.000

S = 8.053, R-Sq = 75.7%, R-Sq(adj) = 75.0%

(a) Write the least-squares regression line, defining your variables. (b) Predict the number of rentals on a day with 14 mph wind. (c) One 14 mph day had 41 rentals. Find and interpret that day's residual. (d) Find the correlation rr. (e) Interpret R-Sq in context. (f) Was the day in part (c) an unusually bad miss for this model? Justify with a number from the output.

Computer regression output: answer key

  1. (a) y^=12.50+6.40x\hat{y} = 12.50 + 6.40x, with xx the firing length in hours and y^\hat{y} the predicted electricity cost in dollars. (b) The intercept is the Coef entry on the Constant row; the slope is the Coef entry on the Hours row. (c) x=0x = 0 is 4 hours below the shortest firing, so the reading is an extrapolation, and a zero-hour firing is not a real firing.

  2. (a) For each additional dog boarded, the predicted daily water use increases by 3.75 gallons, on average. (b) It drops predicted and on average, turning a statement about the fitted line into a promise about individual dogs, and S = 18.41 gallons shows real days miss the line by roughly 18 gallons. (c) 2×3.75=7.502 \times 3.75 = 7.50 more gallons.

  3. (a) y^=15.50+0.28×30=23.90\hat{y} = 15.50 + 0.28 \times 30 = 23.90 degrees Brix. (b) Interpolation, since 30 days lies inside the sampled range of 10 to 40 days. (c) The line gives 32.30 Brix, but 60 days is 20 days beyond the data, so it is an extrapolation and the linear pattern has no support out there.

  4. (a) Predicted 18.00+1.50×40=78.0018.00 + 1.50 \times 40 = 78.00 minutes, so the residual is 70−78.00=−8.0070 - 78.00 = -8.00 minutes. (b) Negative, so the model overpredicted and the load dried 8 minutes faster than the line predicted. (c) No. S = 8.176 minutes is the typical prediction error, and a miss of 8.00 minutes is right at that size.

  5. (a) r=−0.504=−0.71r = -\sqrt{0.504} = -0.71. (b) The sign comes from the negative slope −2.40-2.40, because b=r⋅sysxb = r \cdot \frac{s_y}{s_x} and standard deviations are never negative; R-Sq cannot settle it because 0.710.71 and −0.71-0.71 both square to 0.50410.5041. (c) Yes. SE Coef is always positive, so T=−2.400.4346=−5.52T = \frac{-2.40}{0.4346} = -5.52 carries the slope's sign.

  6. (a) About 88.4 percent of the variation in monthly water use among these 18 buildings is explained by the linear relationship with the number of residents. (b) The other 11.6 percent of that variation is unexplained by the line and comes from other differences between buildings. (c) The first confuses variation with a count of points on the line; the second drops the words the variation in, which is what R-Sq measures.

  7. (a) S = 4.765, in minutes, the units of the response variable. (b) Actual taxi-out times fall about 4.765 minutes away from the predicted times, so that is the typical prediction error. (c) 0.3052 is the standard error of the slope, in minutes per aircraft, describing how much the estimate 2.20 would move across samples, not how far a departure falls from the line. (d) y^=9.50+2.20×8=27.10\hat{y} = 9.50 + 2.20 \times 8 = 27.10 minutes, typically off by about 4.8 minutes.

  8. (a) y^=82.00−3.50x\hat{y} = 82.00 - 3.50x, with xx the midday wind speed in mph and y^\hat{y} the predicted rentals. (b) 82.00−3.50×14=33.0082.00 - 3.50 \times 14 = 33.00 rentals. (c) Residual =41−33.00=+8.00= 41 - 33.00 = +8.00 rentals, so the model underpredicted that day by 8 rentals. (d) r=−0.757=−0.87r = -\sqrt{0.757} = -0.87, negative because the slope is negative. (e) About 75.7 percent of the variation in daily rentals is explained by the linear relationship with midday wind speed. (f) No. S = 8.053 rentals is the typical prediction error, and the residual of 8.00 matches it.

The key gives final answers only. The full worked solution for each problem is on the matching practice set, and every other sheet is on the worksheets page.

More worksheets from this unit