P. 1
Sampling 1

# Sampling 1

|Views: 50|Likes:

Categories:Types, Research

See more
See less

10/24/2012

pdf

text

original

# Sampling and Sample Size Calculation

Lazereto de Mahón, Menorca, Spain September 2006
Sources: -EPIET Introductory course, Thomas Grein, Denis Coulombier, Philippe Sudre, Mike Catchpole, Denise Antona -IDEA Brigitte Helynck, Philippe Malfait, Institut de veille sanitaire Modified: Viviane Bremer, EPIET 2004, Suzanne Cotter 2005, Richard Pebody 2006

Objectives: sampling
 To understand:  Why we use sampling  Definitions in sampling  Sampling errors  Main methods of sampling  Sample size calculation

Why do we use sampling?
Get information from large populations with: ± Reduced costs ± Reduced field time ± Increased accuracy ± Enhanced methods

Definition of sampling

Procedure by which some members of a given population are selected as representatives of the entire population

Definition of sampling terms
Sampling unit (element)  Subject under observation on which information is collected
± Example: children <5 years, hospital discharges, health events«

Sampling fraction  Ratio between sample size and population size
± Example: 100 out of 2000 (5%)

Definition of sampling terms
Sampling frame  List of all the sampling units from which sample is drawn
± Lists: e.g. children < 5 years of age, households, health care units«

Sampling scheme  Method of selecting sampling units from sampling frame
± Randomly, convenience sample«

Survey errors
 Systematic error (or bias)
Sample not typical of population

± Inaccurate response (information bias) ± Selection bias

 Sampling error (random error)

Representativeness (validity)
A sample should accurately reflect distribution of relevant variable in population  Person e.g. age, sex  Place e.g. urban vs. rural  Time e.g. seasonality Representativeness essential to generalise Ensure representativeness before starting, Confirm once completed

Sampling and representativeness

Sampling Population

Sample

Target Population

Target Population

Sampling Population

Sample

Sampling error
 Random difference between sample and population from which sample drawn  Size of error can be measured in probability samples  Expressed as ³standard error´
± of mean, proportion«

 Standard error (or precision) depends upon:
± Size of the sample ± Distribution of character of interest in population

Sampling error
When simple random sample of size µn¶ is selected from population of size N, standard error (s) for population mean or proportion is:

n



p(1-p) n

Used to calculate, 95% confidence intervals Estimated 95% E 2 confidence interval x

X sZ

v s or X s 2 v s x

Quality of a sampling estimate
Precision & validity No precision Precision but no validity

Random error

Systematic error (bias)

Survey errors: example
Measuring height:
 Measuring tape held differently by different investigators loss of precision
± Large standard error
179 178 177 176 175 174 173

 Tape shrunk/wrong systematic error
± Bias (cannot be corrected afterwards)

Types of sampling
 Non-probability samples

 Probability samples

Non probability samples
 Convenience samples (ease of access)  Snowball sampling (friend of friend«.etc.)  Purposive sampling (judgemental)
 You chose who you think should be in the study Probability of being chosen is unknown Cheaper- but unable to generalise, potential for bias

Probability samples
 Random sampling
± Each subject has a known probability of being selected

 Allows application of statistical sampling theory to results to:
± Generalise ± Test hypotheses

Methods used in probability samples
     Simple random sampling Systematic sampling Stratified sampling Multi-stage sampling Cluster sampling

Simple random sampling
 Principle
± Equal chance/probability of drawing each unit

 Procedure
± Take sampling population ± Need listing of all sampling units (³sampling frame´) ± Number all units ± Randomly draw units

Simple random sampling
± Simple ± Sampling error easily measured

± Need complete list of units ± Does not always achieve best representativeness ± Units may be scattered and poorly accessible

Simple random sampling
Example: evaluate the prevalence of tooth decay among 1200 children attending a school     List of children attending the school Children numerated from 1 to 1200 Sample size = 100 children Random sampling of 100 numbers between 1 and 1200 How to randomly select?

EPITABLE: random number listing

EPITABLE: random number listing

Also possible in Excel

Simple random sampling

Systematic sampling
 Principle
± Select sample at regular intervals based on sampling fraction

± Simple ± Sampling error easily measured

± Need complete list of units ± Periodicity

Systematic sampling
 N = 1200, and n = 60   sampling fraction = 1200/60 = 20  List persons from 1 to 1200  Randomly select a number between 1 and 20
(ex : 8)

1st person selected = the 8th on the list   2nd person = 8 + 20 = the 28th etc .....

Systematic sampling

Stratified sampling
 Principle :
± Divide sampling frame into homogeneous subgroups (strata) e.g. age-group, occupation; ± Draw random sample in each strata.

Stratified sampling
± Can acquire information about whole population and individual strata ± Precision increased if variability within strata is less (homogenous) than between strata

± Can be difficult to identify strata ± Loss of precision if small numbers in individual strata
 resolve by sampling proportionate to stratum population

Multiple stage sampling
Principle:  consecutive sampling  example : sampling unit = household
± 1st stage: draw neighborhoods ± 2nd stage: draw buildings ± 3rd stage: draw households

Cluster sampling
 Principle
± Sample units not identified independently but in a group (or ³cluster´) ± Provides logistical advantage.

Cluster sampling
 Principle
± Whole population divided into groups e.g. neighbourhoods ± Random sample taken of these groups (³clusters´) ± Within selected clusters, all units e.g. households included (or random sample of these units)

Example: Cluster sampling
Section 1 Section 2

Section 3

Section 5 Section 4

Cluster sampling
 Advantages ± Simple as complete list of sampling units within population not required ± Less travel/resources required  Disadvantages ± Potential problem is that cluster members are more likely to be alike, than those in another cluster (homogenous)«. ± This ³dependence´ needs to be taken into account in the sample size«.and the analysis (³design effect´)

Selecting a sampling method
 Population to be studied
± Size/geographical distribution ± Heterogeneity with respect to variable

 Availability of list of sampling units  Level of precision required  Resources available

Sample size estimation
 Estimate number needed to  reliably measure factor of interest  detect significant association  Trade-off between study size and resources«.  Sample size determined by various factors:  significance level (³alpha´)  power (³1-beta´)  expected prevalence of factor of interest

Type 1 error
 The probability of finding a difference with our sample compared to population, and there really isn¶t one«.  Known as the (or ³type 1 error´)

 Usually set at 5% (or 0.05)

Type 2 error
 The probability of not finding a difference that actually exists between our sample compared to the population«  Known as the (or ³type 2 error´)

 Power is (1- ) and is usually 80%

A question?
Are the English more intelligent than the Dutch?  H0 Null hypothesis: The English and Dutch have the same mean IQ  Ha Alternative hypothesis: The mean IQ of the English is greater than the Dutch

Type 1 and 2 errors
Decision Reject H0 Accept H0 Truth H0 true H0 false
Type I error Correct decision

Correct decision

Type II error

Power
 The easiest ways to increase power are to:
± increase sample size ± increase desired difference (or effect size) ± decrease significance level desired e.g. 10%

Steps in estimating sample size for descriptive survey
Identify major study variable Determine type of estimate (%, mean, ratio,...) Indicate expected frequency of factor of interest Decide on desired precision of the estimate Decide on acceptable risk that estimate will fall outside its real population value  Adjust for estimated design effect  Adjust for expected response rate     

Sample size for descriptive survey
Simple random / systematic sampling z² * p * q n = -------------d² Cluster sampling z² * p * q n = g* -------------d²
z: alpha risk expressed in z-score p: expected prevalence q: 1 - p d: absolute precision g: design effect

1.96²*0.15*0.85 ---------------------0.03²

= 544

2*1.96²*0.15*0.85 -----------------------0.03²

= 1088

Case-control sample size: issues to consider
    Number of cases Number of controls per case Odds ratio worth detecting Proportion of exposed persons in source population  Desired level of significance ( )  Power of the study (1- )
± to detect at a statistically significant level a particular odds ratio

Case-control: STATCALC Sample size

Case-control: STATCALC Sample size
Risk of alpha error Power Proportion of controls exposed OR to detect 5% 80% 20% >2

Case-control: STATCALC Sample size

Statistical Power of a Case-Control Study
for different control-to-case ratios and odds ratios (with 50 cases)

0

Power

0 04 0 0 4 5 7 0

O O O 4

Control-Case Ratio

Conclusions
 Probability samples are the best  Ensure
± Representativeness ± Precision

 «..within available constraints

Conclusions
 If in doubt«

Call a statistician !!!!

scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->