point_and_interval_estimation_for_population_mean.pdf

Upload: sslbs

Post on 14-Apr-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    1/27

    Point and Interval

    Estimation for PopulationMean

    Ravindra S. Gokhale

    IIM Indore

    1

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    2/27

    Case Leasing

    An auto manufacturer leases cars to small businesses for use in

    visiting clients and other business travel. The contracted lease does

    not specify a mileage limit and instead includes a depreciation fee of

    $0.30 per mile. The contract includes other origination, maintenance,

    and damage fees in addition to the fee that covers the mileage.

    These leases run for a year. A sample of 150 cars (all were a

    particular model of four-door sedan) returned to their dealers

    approximately towards the end of this program averaged 21,000

    miles, with standard deviation s = 2,352 miles. Currently this

    manufacturer has leased approximately 10,000 of these vehicles.

    When the program was launched, the planning budget projected that

    the company would earn (in depreciation fees) $6,500 on average

    per car.

    2Source: Statistics for Business Decision Making and Analysis by Stine and Foster

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    3/27

    Case Leasing

    Motivation:

    Should the manufacturer assume that if it were to check every leased car,

    the average would be 21,000 miles driven?

    Can the manufacturer use a confidence interval to check on the claim of

    $6,500 earnings in depreciation fees?

    Method:

    Are the conditions using a 95% confidence interval for the mean number

    of miles driven per year satisfied?

    Does the method of sampling raise any concerns?

    Can the manufacturer estimate, with a range, the amount it can expect to

    earn in depreciation fees per leased vehicle, on average?

    3Source: Statistics for Business Decision Making and Analysis by Stine and Foster

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    4/27

    Case Leasing

    Mechanics:

    Construct the 95% confidence interval for the number of miles driven per

    year on average for leased cars of this type.

    Construct the 95% confidence interval for earnings over the one-year

    period of the lease, in a form suitable for presentation.

    Message:

    Interpret the 95% confidence interval for the number of miles driven over

    the one year period of the lease.

    Interpret the 95% confidence interval for the average amount earned per

    vehicle. What is the implication fee for the budget claim?

    Communicate a range for the total earnings of this program, assuming

    10,000 vehicles.

    4Source: Statistics for Business Decision Making and Analysis by Stine and Foster

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    5/27

    Case Leasing

    Summary:

    Auto manufacturer leases cars to small businesses.

    For each mile car has travelled, manufacturer gets $0.30.

    Lease runs for one year.

    Manufacturer is expecting to earn $6500 on an average per car.

    Currently the manufacturer has leased approximately 10000 cars

    of a particular model.

    Business Question: Are we going to make profits as expected?

    5

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    6/27

    Case Leasing

    Data in hand:

    A sample of 150 cars returned to the dealers approximately at the

    end of this program averaged 21,000 miles with standard

    deviation 2352 miles.

    6

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    7/27

    Case Leasing

    Statistical Problem:

    What will be the average mileage for all the 10000 cars leased?

    7

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    8/27

    Estimating Population Mean UsingSample Data

    Point Estimation:

    Point estimator is a sample statistic that best describes the

    population parameter.

    e.g. Sample mean is an estimate of population mean

    Notation: An estimator of a parameter is denoted by ^ over the parameter

    symbol

    8

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    9/27

    Case Leasing

    An estimate of the mileage?

    An estimate of the profit?

    Based on the point estimate of the average profit, what do you

    conclude about the average profit from all 10000 cars?

    Is it good to have just one number as the estimate?

    If yes, how? If no, what can be a solution?

    9

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    10/27

    Interval Estimation

    Instead of using one number (point estimate) to describe the

    parameter, we want to use an interval to describe the parameter.

    Such intervals are popularly known as confidence intervals.

    10

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    11/27

    Confidence Interval

    For a given sample, if x-bar is the sample mean of a random

    sample of size n from a normal population with a known

    variance 2, then a 100(1 )% confidence interval on is

    given by:

    where, z/2 is the upper 100(/2) percentage point of the

    standard normal distribution

    11

    nz+x

    nz-x /2/2

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    12/27

    Confidence Interval

    For example, if x-bar is the sample mean of a random sample

    of size n from a normal population with a known variance 2,then a 95% confidence interval on is given by:

    How do you get 1.96?

    12

    n

    1.96+x

    n

    1.96-x

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    13/27

    Confidence Interval (cont)

    Development of a confidence interval

    X-bar ~ Normal(, 2/n)

    Converting it to standard normal distribution,

    Z = [X-bar ] / [ / square-root(n)]

    We write as:

    Re-arranging the terms:

    13

    )(1}zn

    XzP{ /2/2

    )(1}n

    zX

    n

    zXP{ /2/2

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    14/27

    Case Leasing

    Can we assume that X150 follows a Normal Distribution?

    Yes, because of Central Limit Theorem

    Calculate a 95% confidence interval for in the case using X150

    14

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    15/27

    Case Leasing

    How to interpret the Confidence Interval that you have calculated?

    15

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    16/27

    Confidence Interval

    Example of interpreting a 95% Confidence Interval:

    Under repeated sampling, on an average out of 100 confidence

    intervals, 95 will contain the true population parameter.

    16

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    17/27

    Confidence Interval (cont)

    Interpretation of the confidence interval:

    A Confidence Interval is a random interval (depends on the

    content of the sample used to calculate it)

    Wrong interpretation (example):

    The probability that true value of lies within the 95%

    confidence interval is 0.95

    Correct interpretation (example):

    If a large number of random samples are collected and 95%

    confidence interval for is computed from each sample, then

    95% of these intervals will contain the true value of

    17

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    18/27

    Confidence Interval (cont)

    In real life, do you have the luxury to collect infinite number of

    random samples?

    Then how will you interpret it?

    18

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    19/27

    Confidence Interval (cont)

    It is the faith (or the confidence) that you can put in the interval

    that you have calculate.

    Practical interpretation:

    You have a 95% confidence that the interval calculated by

    you contains the true value of the population parameter.

    19

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    20/27

    Confidence Interval (cont)

    Popular choices of the confidence intervals:

    90%, 95%, 99%

    Implications of various percentages?

    Which is better, a wider confidence interval or a narrower

    confidence interval?

    20

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    21/27

    Case Leasing

    What can you conclude?

    21

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    22/27

    Case Leasing

    While calculating the 95% confidence interval for the average

    mileage of the 10000 cars, we used s instead of . Is that

    correct?

    22

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    23/27

    Case Leasing

    While calculating the 95% confidence interval for the average

    mileage of the 10000 cars, we used s instead of . Is that

    correct?

    No

    23

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    24/27

    The t-distribution

    When calculating confidence interval using s instead of, we have

    to use percentiles from t-distribution instead of z-distribution.

    The t-distribution:

    follows a t-distribution with n-1 degrees of freedom.

    What is meaning ofdegrees offreedom?

    Understanding the t-distribution table.

    24

    1-nt~ns-X

    ns

    -X

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    25/27

    Standard Error

    is the standarderror of the estimate

    In general,

    Standard Error in reporting a point estimate:

    The point estimate is a statistic

    The statistic is a random variable and hence subject to variation

    Standard error is the standard deviation of the point estimator

    25

    ns

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    26/27

    The t-distribution

    For a larger sample size (say n > 40), t distribution can be

    approximated well by a normal distribution. Hence for a larger

    sample size, a z-percentile will do a good job irrespective of

    whether sample standard deviation is used for calculating

    confidence interval or population standard deviation is used for

    calculating confidence interval.

    26

  • 7/30/2019 Point_and_Interval_Estimation_for_Population_Mean.pdf

    27/27

    Case Leasing

    While calculating the 95% confidence interval for the average

    mileage of the 10000 cars, we used s instead of . Is that

    correct?

    Since the sample size is 150 which is quite large, the t-distribution can be approximated by a Normal Distribution.

    27