Occurrence Sampling

1
Occurrence Sampling

• Problem how do you know how much time a
  particular person, group, or function is spending
  on any given activity?
• e.g., How much of a students time is spent
  waiting for a report to print in the computer lab
  during peak times?
• How much of the maintenance technicians time is
  spent waiting for repair calls?
• One solution continuous time study
• expensive
• not well suited for nonstandard work
• Alternatively discrete sampling
• select random sample of population
• record activities at discrete intervals

2
Determining Sample Size

• Law of diminishing returns
• amount of information grows proportionately with
  the square root of sample size, n
• cost of information grows directly with n
• therefore, there will be a sample size beyond
  which additional information is not worth the
  cost of additional study
• Sample size depends on
• desired absolute accuracy, A
• note difference between absolute and relative
  accuracy, s
• (estimated) proportion of occurrence, p
• desired confidence level, c

3
Sample size example

• It is estimated that students in the computer lab
  must wait in line for their document to print
  about 45 of the time. To justify an additional
  printer, you wish to verify that estimate within
  15 (relative accuracy) and with a confidence
  level of 90.
• Solution,
• p 0.4
• A (0.45)(0.15) 0.0675
• c 90 ? z 1.64

.0675
-.0675
table 10.1, pg. 137
0.45
0.5175
0.3825
4
Sampling design and data collection

• Overcoming the 3 problems in obtaining a
  representative sample
• Define reasonable strata (categories) for data
  collection
• time of day (morning, afternoon, evening, etc.)
• day of week (or weekend/weekday, week in the
  month, etc.)
• gender
• region
• socio-economic status
• level of education / training
• etc. Base sample size on smallest estimated
  proportion
• Randomness
• defining random sampling times/locations
• randomness with restrictions

table 10.3, pg. 142 (ERGO, Excel)
5
Data Gathering

• Who how?
• person or machine?
• additional duty for employee or hire temp?
• automated data collection?
• level of detail
• the problem of influence
• does the presence of the observer affect the
  actions or performance of the entity being
  observed?
• techniques to minimize influence
• unobtrusive observation
• random sample
• distance, video, etc.
• communication with the observed

6
Data Analysis Use

• Comparing frequency data
• procedure on pg. 145
• Example is there a difference in number of times
  there are students waiting for printouts between
  morning and afternoon?
• na nb 100

7
Frequency example

• Solution,
• 1. Smallest of 4 numbers 25
• 2. Other number in the column 36
• 3. Observed contrast 11
• 4. from Table 10.4, minimum contrast ______
• 5. Compare observed contrast
• Answer Morning is / is not different from
  afternoon.

8
Data Analysis Use

• Other comparison methods
• ?2 (independence) or t-test to test for
  significant difference in means
• control charts to test for time (or sequence)
  effects
• Purpose of the analysis determine if data
  should remain stratified or can be combined
• if no difference, combine data and refer to
  overall proportions
• if there is a difference, keep data, analysis,
  and conclusions separate