Survey Sampling 调查抽样.

INTRODUCTION ·· TO ·· SURVEY SAMPLING February 26, 2003 Karen Foote Retzer Survey Research Laboratory University of Illinois at Chicago www.srl.uic.edu Introduction to Survey Sampling Page 2 of 16 Census or Sample? Census: · Gathering information about every individual in a population Sample: · Selection of a small subset of a population Introduction to Survey Sampling Page 3 of 16 Why Sample instead of taking a Census? · Less expensive · Less time-consuming · More accurate · Some samples can lead to statistical inference about the entire population Introduction to Survey Sampling Page 4 of 16 Probability Sample · Generalize to the entire population · Unbiased results Non-Probability Sample · Exploratory research · Convenience Introduction to Survey Sampling Page 5 of 16 Target Population Definition: The population to which we want to generalize our findings. · Unit of analysis: Individual/Household/City · Geography: State of Illinois/Cook County/ Chicago · Age/Gender · Other variables Introduction to Survey Sampling Page 6 of 16 Examples of Target Populations · Population of adults (18+) in Cook County · UIC faculty, staff, students · Kids under 18 in Cook County Introduction to Survey Sampling Page 7 of 16 Sampling Frame · A complete list of all units, at the first stage of sampling, from which a sample is drawn · Examples: - Lists - Phone numbers in specific area codes - Maps Introduction to Survey Sampling Page 8 of 16 Sampling Frames Example 1: · Population: Adults (18+) in Cook County · Possible Frame: list of phone numbers, list of block maps Example 2: · Population: Females age 40–60 in Chicago · Possible Frame: list of phone numbers, list of block maps Example 3: · Population: Kids under 18 in Cook County · Possible Frame: List of schools Introduction to Survey Sampling Page 9 of 16 Sample Designs for Probability Samples · Simple Random Samples · Systematic Samples · Stratified Samples · Cluster Introduction to Survey Sampling Page 10 of 16 Simple Random Sampling Definition: Every element has the same probability of selection and every combination of elements has the same probability of selection. · Probability of selection: n/N, where n=sample size; N=population size · Use Random Number tables, software packages to generate random numbers · Most precision estimates assume SRS. Introduction to Survey Sampling Page 11 of 16 Systematic Sampling Definition: Every element has the same probability of selection, but not every combination can be selected. · Use when drawing SRS is difficult - List of elements is long & not computerized · Procedure - Determine Population size N & sample size n - Calculate Sampling Interval (N/n) - Pick random start between 1 & Sampling Interval - Take every ith case. · Problem of Periodicity Introduction to Survey Sampling Page 12 of 16 Stratified Sampling: Proportionate · To ensure sample resembles some aspect of population · Population is divided into subgroups (strata) - Students by year in school - Faculty by gender · Simple Random Sample (with same probability of selection) taken from each stratum. Introduction to Survey Sampling Page 13 of 16 Stratified Sampling: Disproportionate · Major use is comparison of subgroups · Population is divided into subgroups (strata) - Compare girls & boys who play Little League - Compare seniors & freshmen who live in dorms · Probability of selection needs to be higher for smaller stratum (girls & seniors) to be able to compare subgroups. · Post-stratification weights Introduction to Survey Sampling Page 14 of 16 Cluster Sampling · Typically used in face-to-face surveys · Population divided into clusters - Schools (earlier example) - Blocks · Reasons for cluster sampling - Reduction in cost - No satisfactory sampling frame available. Introduction to Survey Sampling Page 15 of 16 Determining Sample Size: SRS · Need to consider - Precision - Variation in subject of interest · Formula - Sample size no = CI2 *(pq) Precision - For example: no=1.962 * (.5*.5) .052 · Sample size not dependent on population size. Introduction to Survey Sampling Page 16 of 16 Sample Size: Other Issues · Finite Population Correction n=no/(1+no/N) · Design effects · Analysis of subgroups · Increase size to accommodate non-response · Cost