Transcript Lecture 2
Survey Methodology Sampling EPID 626 Lecture 2 1/19/00 What is sampling? • Population: The collection of all possible measurements that could be used to address the study question. • Sample: (v.) To select a small subset of a population representative of the whole population. (Fowler, 1993) 1/19/00 • It is assumed that only chance could cause the composition of the sample to differ from the composition of the population in all aspects other than the quantity of data contained in each. (Hirsch and Riegelman, 1995) 1/19/00 • The key to good sampling is finding a way to give all (or nearly all) population members the same (or a known) chance of being sampled, and to use probability methods for choosing the sample. (Fowler, 1993) 1/19/00 Critical sampling issues • Whether or not to use a probability sample • The sample frame (those who actually have a chance to be sampled) • The size of the sample 1/19/00 Critical sampling issues (con’t) • The sample design (the particular strategy used for sampling people or household) • The rate of response (the percentage of those sampled for whom data are actually collected) (Fowler, 1995) 1/19/00 Sample frame • The set of people that has a chance to be selected, given the sampling approach that is chosen. • Question: How well does the sample frame correspond to the population you want to describe? (Fowler, 1993) 1/19/00 Examples of sampling frames • List of registered drivers in Louisiana • List of patients who have been treated at a clinic in the past year • Greater New Orleans residential phone listing • List of all public schools in Virginia 1/19/00 Here is our sampling scenario • Population: Roosevelt High School students N=99 • Sampling frame: List of students, numbered 01-99 • Desired sample size: n=33 1/19/00 Sampling strategies • One-stage sampling – Simple random sampling – Systematic sampling – Stratified sampling • Multi-stage sampling – Area probability sampling 1/19/00 Simple random sampling • Each member of the study population has an equal probability of being selected. • Analogous to drawing a number from a hat. • Each sample is sampled from the sampling frame one at a time, independent of one another, and without replacement. 1/19/00 Simple random sampling • We do it by numbering the sample frame, then using a computer, a table of random numbers, or another random generator to randomly choose observations from the list. 1/19/00 Systematic random sample strategy • Each member of the study population is listed, a random start is designated, then members of the population are selected at equal intervals. (Henry, 1990) 1/19/00 Systematic random sampling • Determine your interval: i=N/n • Select a random start between 0 and i • Select every ith person • Cautionary note about ordered lists 1/19/00 Roosevelt systematic random sampling • • • • i=99/33=3 (Round down if i is not an integer) Select a random start from 1 to 3 Select every 3rd student from the random start • So if start is 2, select 2, 5, 8 etc. 1/19/00 Stratified sampling strategy • Each member of the study population is assigned to a group or stratum, then a simple or systematic random sample is selected from each stratum. • This reduces normal sampling variation and ensures that the sample reflects the total population with regard to the stratifying variable. 1/19/00 Stratified Disproportionate Sampling • Can oversample a stratum with high variability to increase the precision of an estimate • Oversample a particular stratum to increase the n for the subpopulation without a corresponding increase in the total N. • Important to weight data accordingly for analysis 1/19/00 Disproportionate Sampling White Af.Am. N population 4,500 500 % of population 90 10 1/i 1/10 1/5 n 450 100 % of sample 81.8 18.2 Weight 1 ½ Weighted n 450 50 Weighted % 90 10 1/19/00