1 Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc. Chapter 15 Sampling
1Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Chapter 15
Sampling
2Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sampling
Media announcements drive understanding of sampling: A random sample showed . . . Two of three people chose . . . Preferred two to one over . . .
However, in most instances, television, newspapers, and advertisements do not explain their sampling techniques
3Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sampling Theory
Sampling is selecting A sampling plan is the researcher’s plan for selection A sample includes all elements included in a study The way a sample is chosen determines how it can
be generalized Developed to determine mathematically how to
acquire a representative sample Derived from survey research first used for census
reporting
4Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Key Concepts of Sampling Theory
Populations and elements Sampling or eligibility criteria Sample representativeness Sampling error Randomization Sampling frames Sampling plans
5Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Populations and Elements
6Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Populations and Elements (Cont’d)
Population Characteristic Target population Elements Accessible population Subject, participant, informant Generalizing Hypothetical population
7Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sampling or Eligibility Criteria
Sampling criteria Inclusion criteria Exclusion criteria Twiss study If federally funded, study should be inclusive
of underrepresented groups Should not use them exclusively, especially when
risk is present Should not exclude them totally, especially when
benefit is likely
8Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sample Representativeness
As much like the target population as possible In relation to the variables under consideration In relation to possibly extraneous variables
Accessible population must be representative of the target population
Population parameter Precision
10Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sampling Error (Cont’d)
Any sample’s mean is expected to be slightly different from the population mean
Occurs because of variation—random or systematic The difference between a sample statistic and the
true population parameter The larger the sampling error, the less representative
the sample is of the target population The larger the sample, the more representative the
sample is of the target population Sampling error decreases power
11Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Random Variation
Expected, normal difference from the mean in individual values
The values are randomly scattered around the mean
As sample size increases, sample mean is more likely to have a value similar to that of the population mean
12Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Systematic Variation/Systematic Bias
All the values in the sample may tend to be higher or lower than the population’s mean
Occurs when the sample is NOT representative of the whole population
Nonrandom samples have more systematic bias
Exclusion criteria and high refusal rates increase systematic bias
Issues with generalizability
13Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Refusal Rate and Attrition
Typically, refusal to participate eliminates people with LESS of something
Sample attrition—withdrawal or loss of subjects from a study
14Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Randomization
Random selection—everyone in the accessible population has an equal chance of being selected
Random assignment—the sample is randomly assigned to experimental (treatment) group or control group
If subjects are not randomly assigned to a group, the group not receiving the treatment is called a comparison (not control) group
15Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sampling Frame and Sampling Plan
Sampling frame—list of all members in the accessible population asked to participate
Sampling plan—strategies for obtaining the sample probability (random) nonprobability (nonrandom) methods
16Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Probability (Random)Sampling Methods
Reduce sampling error Leaves the selection to chance Increases study validity Four sampling designs
Simple random sampling Stratified random sampling Cluster sampling Systematic sampling
17Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Simple Random Sampling
Most basic Elements are selected at random from the
sampling frame With replacement Without replacement
18Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Stratified Random Sampling
Used when researcher wants to include certain population variables so that none is unrepresented
Disproportionate sampling Proportionate sampling
19Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Cluster Sampling
Samples according to natural clusters Useful when
The population is heterogeneous Elements of population cannot easily be identified
One stage or multistage Provides a means for obtaining a larger
sample at a lower cost Some disadvantages
20Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Systematic Sampling
Useful when a master (unordered) list exists Involves selecting every kth individual on the
list, using a starting point selected randomly
21Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Nonprobability (nonrandom) Sampling Methods
Not every element of the population need have an equal opportunity to be included
More likely to be nonrepresentative of the population
Five types used in nursing: convenience, quota, purposive or purposeful, network or snowball, and theoretical
22Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Convenience Sampling
Many quantitative/some qualitative research “Accidental sampling” Right place, right time May or may not represent the population the
researcher is targeting Considered a weak approach—biases
23Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Quota Sampling
Many quantitative/some qualitative research Convenience sampling with set proportions
from certain groups (just like stratified random sampling, but NOT random)
Used to assure representation May be proportionate or disproportionate in
design
24Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Purposive Sampling
Certain elements purposefully selected Sometimes referred to as purposeful,
judgmental, or selective sampling Conscious selection of the persons most
likely to provide useful information Commonly used in qualitative research;
provides understanding of a specific topic Less commonly used in quantitative research
25Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Network (Snowball) Sampling
Used to obtain subjects difficult to identify or connect with in other ways
More common in qualitative research, but may be used in quantitative research
Sometimes called snowball sampling, and occasionally chain sampling
Takes advantage of social networks
26Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Theoretical Sampling
Used exclusively in qualitative research Most common in grounded theory research Allows researcher to select participants who
are able to contribute to a theory in progress
27Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sample Size in Quantitative Research
Power Power analysis Factors that affect power:
Effect size Type of study Number of variables Sensitivity of the measurement methods Data analysis techniques
28Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Effect Size (ES)
Extent to which a phenomenon is present in a population
As ES decreases, power decreases To demonstrate a small ES, the researcher
requires a larger sample To demonstrate a large ES, the researcher can
get by with a smaller sample ESs vary according to the population being
studied
29Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Type of Study
Descriptive case studies—small samples Other descriptive studies and correlational—
often large samples More variables require bigger samples Quasi-experimental and experimental
Small, in past—just enough to meet power analysis needs
Some multi-site moderate samples Confirmatory and exploratory studies
30Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Number of Variables
As number of variables under study grows, the needed sample size may also increase
If variables are highly correlated with dependent variable, effect size will increase, and sample size can be reduced
Variables included in data analysis must be carefully selected
31Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Measurement Sensitivity
Precision Range of measured values influences power
ESs vary according to how near the value is to the population mean
The wider the range of values sampled, the larger the ES
32Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Data Analysis Techniques
Differ by test Parametric tests more sensitive than non-
parametric tests, when used appropriately For t-test and ANOVA, power is increased
with equal group sizes For unequal groups, total sample size must
be larger Chi-square (2) test—weakest
33Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Sample Size in Qualitative Research
Focus is on the quality of information obtained
Saturation of data Scope of the study Nature of the topic Quality of the data Study design
34Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Research Settings
Natural setting (or field setting) Partially controlled setting Highly controlled setting (laboratory)
35Copyright © 2013, 2009, 2005, 2001, 1997 by Saunders, an imprint of Elsevier Inc.
Recruiting and Retaining Research Participants
Recruitment strategies differ, depending on type of study and setting
Often wise to over-enroll, to compensate for attrition
May offer a small remuneration Retention rates absolutely impossible to
predict