Transcript Introduction to Decision Analysis

Introduction to
Sampling
(Dr. Monticino)
Assignment Sheet
Read Chapter 19 carefully
Quiz # 10 over Chapter 19
Assignment # 12 (Due Monday April 25th)
Chapter 19
 Exercise Set A: 1-6,8,11
Overview
Language of statistics
Obtaining a sample
Statistical Terms
Population
 The whole class of individuals of interest
 Voters
 Customers
 Marbles in a box
Parameter
 Numerical facts about the population
 Percentage who will vote for candidate A
 Average income
 Proportion of white marbles
Statistical Terms
Sample
Part of a population
 1000 eligible voters called at random
 First 400 customers on Tuesday morning
 5 marbles drawn from the box with
replacement
Statistic
Numerical value obtained from sample
used to estimate population parameter
Sampling
Generally, determining population
parameters by studying the whole
population is impractical
Thus, inferences about population
parameters are made from sample
statistics
This requires that the sample represent
the population
Sampling
To obtain a representative sample,
probability methods are used
Employ an objective chance process to pick
the sample
 No discretion is left to the interviewer
The probability of any particular
individual in the population being selected
in the sample can be computed
Simple Random Sampling
Most straightforward sampling method is
simple random sampling
 Individuals in the sample are drawn at random
from the population without replacement
 Each individual is equally likely to be selected and
each possible subset of individuals is equally
likely to be selected
 Care must be taken to ensure that the selection
process is not biased
Other Sampling Techniques
Multi-stage cluster sampling
Other Sampling Techniques
Quota sampling
Sample is hand-picked to resemble the
population with respect to selected key
characteristics
 Selection bias
 Response/Non-response bias
Good and Bad Samples
Samples obtained by probability methods
give a good representation of the population
 In theory, simple random sampling gives best
representation
Cluster samples, properly weighted, provide
reasonable compromise between representing
population and practical issues
Good and Bad Samples
Quota samples typically introduce
selection and response/non-response
bias
Samples of convenience rarely
represent the population. Avoid these
When a sampling procedure is biased,
taking a larger sample does not help
Good and Bad Samples
When examining a sample survey, ask:
What is the population?
What is the parameter being estimated?
How was the sample chosen?
What was the response rate?
Address these same questions when
designing a sampling procedure
Sampling Error
Even a well designed sampling procedure
may result in an estimate which differs from
the true value of the population parameter
 Bias
 Chance error
It is important to have a measure of the
sampling error of the parameter estimate
(Dr. Monticino)