5.

Onderzoeksopzet: Populaties, steekproeven

Objectives

You can:

 Distinguish between a population and a sample (Ch.13)

 Distinguish between a target population and an accessible population (Ch.13)
 Describe the steps in the process from defining the target population up to arriving
at the actual study sample (Ch.13)
 Describe the purpose of inclusion and exclusion criteria in sampling for research
studies (Ch.13)
 Describe the concepts of sampling error and sampling bias (Ch.13)
 Describe probability and nonprobability sampling, their subtypes and possible
sample techniques (Ch.13)
 Discuss issues in recruiting an adequate sample size (Ch.13)
 Populations, Samples and the Sampling process
 Population - the aggregate of persons or objects that meet a specified set of criteria and to whom we wish
to generalize results of a study
 Sample - subgroup of the population serving as the reference group for estimating characteristics of and
drawing conclusions about the population (inferential statistics).

 Target population
 Overall group to which findings will be generalized
 Accessible population
Variation?  more / less in  Persons who have an actual chance to be selected
…………………..than in………………………  Sample
Handbook: Chapter 13 – Choosing a sample  Derived from the accessible population

How do we chose a sample?

 Identify the target population
 Specify inclusion and exclusion criteria
 Inclusion criteria
 Exclusion criteria
 Potential variables that may confound the results or interfere with interpretation of the findings
 Exclusion criteria for RCTs
 E.g. comorbidities / health or demographic traits that could confound (vertekenen) examination of
relationships among the study variables
 E.g. medical condition that would put participants at risk with the experimental treatment
 E.g. not being able to understand study materials

What research type has very strict in- and exclusion criteria?..........................
 How is the sampling process presented?

http://www.consort-statement.org/consort-statement/flow-diagram

HIIT voor cardiorespiratoire fitness?

Onderzoeksvraag Niet directioneel:
Directioneel:

H0
H1
Inclusie criteria

Exclusiecriteria

Steekproeftrekking

https://pubmed.ncbi.nlm.nih.gov/30509862/
 How many participants is enough?
 Number of participants = Sample Size
 The more the better?
 As many as we can pay for?
 As many as are willing to participate?

 We can do an educated guess

 ………………………………………. – technique to estimate the number of participants
 …………………......................- Ability to find significant effects when they exist (Onderscheidend
vermogen, probabiliteit om de nulhypothese af te wijzen als deze onjuist is)
 Depends on:
 Variability in the sample
 Anticipated magnitude of the effect
 Small sample?  low power  more difficult to find significant effects

Studiehulp – Steekproeftrekking
• Beschrijving Voordelen? Nadelen?
• Probability? Non Probability?

Simple

Convenience

Systematic

Cluster

Stratified

Dr Nic’s Maths and Stats - Sampling: Simple Random, Convenience, systematic, cluster, stratified - Statistics Help  https://www.youtube.com/watch?v=be9e-Q-jC-
0&list=PLm9FYjKtq7PyqaxpVkODL_PramidLWPcB
 How to select a representative sample?
 A good sampling procedure can minimize the degree of bias/error
 Recruitment plan includes method for sampling
 1) Probability Sampling versus 2) Non probability sampling

 Probability sampling aka Random sampling - every member of a

population has an equal probability of being selected
 Results in random sample, advantages:
 1) free of bias;
 2) considered representative of the population
 Non probability sampling – using nonrandom sampling methods
 More often used in clinical research
 Consequences - outcomes require some caution for generalization

Probability sampling – vul in

 ……………………………………………………..sampling
  Each member of the accessible population has an equal chance of being selected
 ……………………………………………………..sampling
  Persons are randomly chosen from unordered lists using fixed sampling intervals (e.g.
every 10th person). The interval is detemined by desired sample size and length of the list.
This is a random process as long as the list has an unbiased order
 ……………………………………………………..sampling
  Based on a characteristic of interest, the population is divided into groups (strata); random
samples are drawn for each group so that the sample represents the target population
 ……………………………………………………..sampling
  Random sampling at multiple stages, e.g. provinces – cities – schools – classes – students;
for very large populations
 ……………………………………………………..sampling
  Used when certain strata are underrepresented, resulting in underrepresentation;
Oversampling of these strata for stronger representation; use of weights to adjust data for bias

Kies uit: Cluster – Disproportional - Simple random – Stratified random - Systematic -

 Sampling procedure?

?..…………………

Non probability sampling – vul in

 ……………………………………………………..sampling
  Subjects are chosen on the basis of their availability
 e.g. they get a treatment in a certain practice, they enter a gym, they volunteer
 ……………………………………………………..sampling
  Stratified sampling, but not random
 Subject are recruited to represent strata in proportion to their number in the population
 ……………………………………………………..sampling
  Subjects are hand picked and invited because of known characteristics
 ……………………………………………………..sampling
  Small numbers are purposively recruited; They help to identify other potential
participant’; Used when subjects are difficult to identify, e.g. dealing with sensitive topics

Kies uit: Convenience - Purposive - Quota - Snowball

 !Sampling error aka Sampling variation!
 We cant measure the total population  We recruit and measure a sample  We
measure the variables of interest and summarize responses from the sample (Centrum en
Spreidingsmaten)
 Just by chance, the sample average is very likely different from the population’s average
  Sampling error aka Sampling variation - Difference between sample values and
population values
 ‘In most situations, we don’t actually know the population parameters. We use statistical
procedures to estimate them from our sample statistics, always understanding there will
likely be some error.’
 The essence of random sampling is that these sampling differences are due to chance and
are not a function of any human bias, conscious or unconscious.’

Handbook

!Sampling bias !
 Sampling bias: ……………………………………………………………………………………………..
 Results in variability, not by chance
 1. Conscious sampling bias
 Sample is selected purposefully (doelbewust)
 E.g. clinician chooses patients with minimal dysfunction to demonstrate a treatment’s
effectiveness, eliminating those subjects who were not likely to improve
 2. Unconscious sampling bias
 E.g street interviews
 Conclusions are not useful for describing attitudes of the “general public”
 Do you want to make valid generalizations from a sample to the population?
 Validity depends on method of selecting subjects
 A larger sample does not necessarily control for bias!
 Understanding confidence intervals
 Sample:………………………………………………………………………………………………………..
 Inference:……………………………………………………………………………………………………..
 Sampling error – Sampling variation:…………………………………………………………….
 Confidence Interval (CI):…………………………………………………………..…………………..
 Width CI affected by:
 1. …………………………………………………………………………..
 When is the confidence interval smaller & why?  ………………………………………………………………......……..
 When is the confidence interval wider & why?  ……………………………………………………………………………..
 2. …………………………………………………………………………..
 When is the confidence interval smaller & why?  ………………………………………………………………......……..
 When is the confidence interval wider & why?  ……………………………………………………………………………..
 You do not need to know the ways to calculate CIs as explained in this video

Dr Nic’s Maths and Stats - Understanding Confidence Intervals: Statistics Help

vwM5 - Sampling error and variation in statistics and data science

Dr Nic’s Maths and Stats - Sampling error and variation in statistics and data science (6.18 min)