Multi-Focus on Practices 2, 3, and 4

Synthesize sampling, probability, and regression: FRQ

3 versions of one AP Statistics free-response question, each with a 10-point rubric and a model answer for every point. Write your response on paper before you open the rubric.

These are StatsLearn originals written in the AP style, not released exam questions. Topics: 1.11 Random Sampling, 2.3 Estimating Probabilities Using Simulation, 2.4 Introduction to Probability, 4.1 Sampling Distributions for Sample Means, 5.3 Linear Regression Models, 5.4 Residuals.

Version 1

An environmental lab has 900 field sensors. It selects 45 sensors by a simple random sample. Each sensor has independent failure probability 0.12 in a storm. Maintenance time y is modeled from sensor age x by y-hat = 9 + 1.8x. Maintenance-time population standard deviation is 12 minutes.

  1. (a) Describe the sampling method and its scope of inference.
  2. (b) For a set of 4 sensors, calculate the exact probability of at least one failure and compare it with a simulation that found 204 successes in 500 trials.
  3. (c) For a sensor aged 6 years with observed maintenance time 22.0 minutes, compute and interpret its prediction and residual.
  4. (d) Find the standard deviation of x-bar for samples of 36, then explain the limit on a causal reading of the regression.
Show the rubric and model answers

Part (a): 2 points

  • Describes a simple random sample. Label the 900 sensors and randomly select 45 distinct labels so every group of 45 has an equal chance.
  • States the generalization supported by random sampling. Results from the simple random sample can be generalized to the lab's field sensors, subject to nonresponse or missing measurements.

Part (b): 3 points

  • Computes the exact probability of at least one failure. P(at least one) = 1 - 0.88^4 = 0.4003.
  • Computes the simulation estimate. The simulated probability is 204/500 = 0.4080.
  • Compares simulation with the exact probability. The simulation estimate is close to the exact probability, and a small gap is expected from simulation variation.

Part (c): 3 points

  • Computes predicted maintenance time. At age 6, y-hat = 9 + 1.8(6) = 19.8 minutes.
  • Computes the residual. Residual = 22.0 - 19.8 = 2.2 minutes.
  • Interprets the residual's direction. This sensor took longer to maintain than the model predicted.

Part (d): 2 points

  • Computes the sampling distribution spread for x-bar. SD(x-bar) = 12/sqrt(36) = 2.000 minutes.
  • Avoids a causal claim from the regression. The regression describes an association between sensor age and maintenance time. Age was not randomly assigned, so it does not establish causation.

Version 2

An environmental lab has 1180 field sensors. It selects 57 sensors by a simple random sample. Each sensor has independent failure probability 0.16 in a storm. Maintenance time y is modeled from sensor age x by y-hat = 13 + 2.2x. Maintenance-time population standard deviation is 16 minutes.

  1. (a) Describe the sampling method and its scope of inference.
  2. (b) For a set of 5 sensors, calculate the exact probability of at least one failure and compare it with a simulation that found 389 successes in 660 trials.
  3. (c) For a sensor aged 10 years with observed maintenance time 37.2 minutes, compute and interpret its prediction and residual.
  4. (d) Find the standard deviation of x-bar for samples of 52, then explain the limit on a causal reading of the regression.
Show the rubric and model answers

Part (a): 2 points

  • Describes a simple random sample. Label the 1180 sensors and randomly select 57 distinct labels so every group of 57 has an equal chance.
  • States the generalization supported by random sampling. Results from the simple random sample can be generalized to the lab's field sensors, subject to nonresponse or missing measurements.

Part (b): 3 points

  • Computes the exact probability of at least one failure. P(at least one) = 1 - 0.84^5 = 0.5818.
  • Computes the simulation estimate. The simulated probability is 389/660 = 0.5894.
  • Compares simulation with the exact probability. The simulation estimate is close to the exact probability, and a small gap is expected from simulation variation.

Part (c): 3 points

  • Computes predicted maintenance time. At age 10, y-hat = 13 + 2.2(10) = 35.0 minutes.
  • Computes the residual. Residual = 37.2 - 35.0 = 2.2 minutes.
  • Interprets the residual's direction. This sensor took longer to maintain than the model predicted.

Part (d): 2 points

  • Computes the sampling distribution spread for x-bar. SD(x-bar) = 16/sqrt(52) = 2.219 minutes.
  • Avoids a causal claim from the regression. The regression describes an association between sensor age and maintenance time. Age was not randomly assigned, so it does not establish causation.

Version 3

An environmental lab has 1460 field sensors. It selects 69 sensors by a simple random sample. Each sensor has independent failure probability 0.20 in a storm. Maintenance time y is modeled from sensor age x by y-hat = 17 + 2.6x. Maintenance-time population standard deviation is 20 minutes.

  1. (a) Describe the sampling method and its scope of inference.
  2. (b) For a set of 6 sensors, calculate the exact probability of at least one failure and compare it with a simulation that found 612 successes in 820 trials.
  3. (c) For a sensor aged 14 years with observed maintenance time 55.6 minutes, compute and interpret its prediction and residual.
  4. (d) Find the standard deviation of x-bar for samples of 68, then explain the limit on a causal reading of the regression.
Show the rubric and model answers

Part (a): 2 points

  • Describes a simple random sample. Label the 1460 sensors and randomly select 69 distinct labels so every group of 69 has an equal chance.
  • States the generalization supported by random sampling. Results from the simple random sample can be generalized to the lab's field sensors, subject to nonresponse or missing measurements.

Part (b): 3 points

  • Computes the exact probability of at least one failure. P(at least one) = 1 - 0.80^6 = 0.7379.
  • Computes the simulation estimate. The simulated probability is 612/820 = 0.7463.
  • Compares simulation with the exact probability. The simulation estimate is close to the exact probability, and a small gap is expected from simulation variation.

Part (c): 3 points

  • Computes predicted maintenance time. At age 14, y-hat = 17 + 2.6(14) = 53.4 minutes.
  • Computes the residual. Residual = 55.6 - 53.4 = 2.2 minutes.
  • Interprets the residual's direction. This sensor took longer to maintain than the model predicted.

Part (d): 2 points

  • Computes the sampling distribution spread for x-bar. SD(x-bar) = 20/sqrt(68) = 2.425 minutes.
  • Avoids a causal claim from the regression. The regression describes an association between sensor age and maintenance time. Age was not randomly assigned, so it does not establish causation.

Keep going