Section 1-4
Collecting Sample Data
DEFINITIONS
•Observational Study: observing andmeasuring specific characteristics withoutattempting to modify the subjects beingstudied
•Experiment: apply some treatment andthen observe its effects on the subjects
TYPES OF OBSERVATIONALSTUDIES
•In a cross-sectional study, data areobserved, measured, and collected at onetime.
•In a retrospective (or case-control) study,data are collected from the past by goingback in time (through examination ofrecords, interviews, etc.).
•In a prospective (or longitudinal orcohort) study, data are collected in thefuture from groups (called cohorts) sharingcommon factors.
DESIGNING AN EXPERIMENT
1.Identify your objective
2.Collect sample data
3.Use a Random Procedure that avoids bias
4.Analyze the data and form conclusions
CONFOUNDING
Confounding occurs in an experiment whenthe experimenter is not able to distinguishbetween the effects of different factors.
SAMPLING
•Random and Simple Random
•Systematic
•Convenience
•Stratified
•Cluster
Sampling refers to the method used to selectthe subjects being studied. The methods ofsampling we will study are:
RANDOM SAMPLING
In a random sample members of thepopulation are selected in such a way that eachindividual member has an equal chance ofbeing selected.
SIMPLE RANDOM SAMPLING
A simple random sample of size n isselected in such a way that every possiblesample of the same size n has the samechance of being chosen.
EXAMPLE
Suppose there is a class with 5 men in one rowand 5 women in the other. I can get a randomsample of size 5 by flipping a coin. If heads, Iselect the row of women. If tails, I select therow of men. Each student has an equal chanceof being selected, and the sample is clearlyrandom. However, it is not an simple randomsample of size 5 because not every group of 5students is equally likely to be selected. Withthe above method, I cannot get 2 men and 3women.
SYSTEMATIC SAMPLING
Systematic sampling: randomly select somestarting point and then select every kthelement.
CONVENIENCE SAMPLING
Convenience Sampling: use results that arereadily available.
STRATIFIED SAMPLING
Stratified Sampling: subdivide the populationinto subgroups that share the same characteristic,then draw a sample from each stratum.
CLUSTER SAMPLING
Cluster Sampling: subdivide the population intosections (clusters), randomly select some of theclusters, then choose ALL the members from theselected clusters.
ERROR
•Sampling Error (or random sampling error):the difference between a sample result and thetrue population result; such an error results fromchance sample fluctuations.
•Nonsampling Error: sample data that areincorrectly collected, recorded, or analyzed (suchas by selecting a biased sample, using a defectiveinstrument, or copying the data incorrectly).
•Nonrandom Sampling Error: the result of usinga sampling method that is not random, such asusing a convenience sample or a voluntaryresponse sample.