Sampling

1
Sampling

• This lecture ties into Terre Blanche chapter 15
• Now you have a design, and a protocol plus
  measures who do you run it on?
• Selecting a group of people is called sampling
  many strategies

2
Samples Populations

• Population All possible observations of a
  particular variable
• Eg all people on earth (past, present future)
• Cannot experiment on populations!
• Sample subset of a population selected to
  estimate the behaviour of the population
• Looking at a few people, we can guess what is
  happening in the group as a whole

3
Sample Ex-sample
What percentage of these dots are black and which
are white?
4
Sample example
In this small sub section (the sample), There
are 53 black, 47 white
In the big picture, there are in fact 59 black,
41 white
Even though we used a sample, we approximated the
big picture well!!
5
Two big questions about samples

• How many people (n?)
• Depends on which analysis tool you use
• Calculating statistical power will tell you
• About 20 per group in the design is good
• Who to use
• Depends on the population you want to speak of
  (pretty complicated)

6
Importance of sampling properly

• A sample exists to represent its parent
  population
• We must know what the actual parent population
  is, otherwise we draw false conclusions
• If we sample only women, we cannot safely make
  claims about men
• Eg. Student sampling problem is psychology
  about people or about students?

7
Deciding who to choose

• Two basic sampling methods
• Probability sampling
• Each member of the population has a certain
  probability of being selected
• Non-probability sampling
• Members selected not by mathematical rules, but
  by other means (eg. Conveninece, access)

8
Probability sampling strategies

• There are several different ways of doing
  probability sampling
• We look at 3
• Systematic sampling
• Random sampling
• Stratified sampling

9
Systematic sampling

• Put all your population on a list
• Select every nth subject (eg every 12th)
• n is determined by desired sample size
• Eg. With a population of 300, if we want a sample
  of 10, choose every 30th case
• Only useful if you have a complete list of the
  population
• Eg classlist customer database

10
Random sampling

• Random without a rule or method
• Assign each person a probability of being chosen
• Try to match the ratios that exist in the
  population
• Eg if you suspect 60 males, 40 females, then
  assign those odds to selecting the next person

11
Stratified sampling

• Expands on random sampling
• Build sub categories, then sample randomly inside
  each one
• Eg decide you will have 10 men and 10 women
• Random sampling cannot ensure equal group size
  stratification can

12
Non-probability sampling

• Methods used when distribution of the sample is
  not important
• Used when sampling frame is not known
• Cannot be used for most statistical analyses
• Well suited for qualitative research, where
  distribution is not important

13
NP sampling strategies

• Haphazard sampling use the first people at hand
• Most convenient method, cheapest
• Quota sampling Stratified haphazard sampling
• Like haphazard, but ensures subpopulations are
  included

14
NP sampling strategies (2)

• Purposive sampling Use expert judges to identify
  candidates, select those
• Very rare populations
• Snow-ball sampling Recruit one subject, that
  subject identifies others, who identify others
• Used for covert subpopulations, non-cooperative
  groups