Occurrence Sampling: Ch' 10

1
Occurrence Sampling Ch. 10
2
Occurrence Sampling

• Address the Problem
• How Many Observations Are Needed ?
• What is a Representative Sample ?
• How to Collect Data
• Data Analysis
• Applications of Results Keep Management Informed

3
What is Occurrence Sampling?

• Taking data at discrete points in time
• Gaps exist between samples
• Assumes binomial distribution
• Yes/no
• Working/Not working
• Also called Work Sampling

4
What is Sampled?

• Delays Is equipment being used?
• Work-Task Ratios How much time is spent on
  select activity?
• Status of Output Is load damaged vs. not
  damaged?
• Other binomial data on equipment, workers, etc.

5
Underlying Population

• Need to have an idea of the distribution of the
  population, relative frequency histograms work
  well
• This profile of the population should be
  considered for all key characteristics
• Age, Sex, Other demographics
• Time of day events
• The more thorough this background is, the better
  the sampling process will represent a true
  picture
• If doing simple correlation, then this data can
  help stratification analysis

6
Selecting the Sample

• Information costs money, how much to buy?
• Too little gt poor estimates
• Too much gt waste of time, money
• Determining the precision of the information
  will help in choosing sample size
• Cluster sampling If a profile of the population
  is available, it might be possible to choose a
  subset of the entire population and select sample
  from there
• Systematic sampling involves selecting candidate
  element from a list at regular intervals (i.e.,
  every 5th)

7
Quota Sampling

• Until 1948, was used in presidential election
  polls
• General information given as to who to select,
  but final selection of the respondents is left up
  to the subjective judgment of the interviewer
  rather than being chosen objectively
• Can result in unintentional segmentation around
  factors not identified as stratifying items
• example Find two men and three women, make sure
  one is under 25 years old, the rest over 25.
• May end up selecting people from a better part of
  town
• More attractive people selected

8
Probability Sampling

• Improved accuracy with smaller sample size
• Randomization performed so that quotas for
  certain variable can be selected, but others will
  not influence results
• Uncontrolled and uncontrollable factors are
  balanced out
• Use systematic sampling scheme, either manually
  or through computer generation
• What are surveys at a shopping mall?

9
Comparison of Quota and Probability Sampling

• Washington State Poll of 1948
• Actual Vote Probability Quota
• Dewey 42.7 46.0 52.0
• Truman 52.6 50.5 45.3
• Wallace 3.5 2.9 2.5
• Quota sampling resulted in incorrect winner
  predicted.
• Races prior to 1948 were not quite so close, so
  quota sampling predictions were correct.

10
Sample Size Requirements

• To determine the numbers of samples needed, you
  first have to know the desired accuracy needed
• More data does not mean better information (law
  of diminishing returns)
• Key terms
• A Desired absolute accuracy
• p Proportion of occurrence
• c Confidence level desired

11
Calculating Sample Size

• sp
• where p occurrence proportion
• And n number of observations
• We assume normality of sample size, since ngt30
• Use Table 10.1 for Z-table Confidence levels
• Switch to absolute accuracy (from relative
  accuracy) for calculations
• Again, use confidence levels to determine
  long-run probability

12
Sampling Procedure Problems

• Your goal is to get representative data!
• Be aware of and avoid
• Stratification (or mixing of disparate groups)
• Influence of Sampling on the people being sampled
  (Hawthorn effect)
• Periodicity (non-random events)

13
How to Collect Samples

• Samples can be taken
• Randomly (no pattern)
• Randomly, with restrictions (allowing
  stratification, like separating morning unloading
  of goods and afternoon packing of final product)
• Periodically (on scheduled basis, when bias can
  be avoided)

14
How to Collect Samples (cont.)

• Can collect results with
• clipboard sampling (manual)
• computer sampling (on-line sampling, such as
  using a passive monitoring to find out how many
  people are logged on at a particular time)
• automated systems (like Palm Pilots, Dolphin
  equipment, etc.)