Unformatted text preview: JS202: Research Design, Methods, and Evaluation Data Collection The value of research depends heavily upon how the data were collected What is an outcome measure? Brainstorm in advance about all possible effects Participant characteristics (eg) Age Sex SES Race/ethnicity Attitude towards Px Severity of problem Motivation for participation Expectations Degree of family support Ways of collecting data Formal interviews (phone/mail) Program records Agency files Psychometric tests Diaries Clinical examinations Physical evidence Beware of social desirability and telescoping (and the original purpose of the data) Sampling To avoid bias, we need to find a group that is representative of the larger population that we are trying to describe Sampling can occur through: Asking questions Making observations Examining written records Sampling terms Census Research population Element Sampling frame Sample size Sampling error Confidence level Confidence interval Sampling techniques Random/probability sampling Systematic sampling Stratified/proportionate sampling Disproportionate sampling Cluster sampling Multistage sampling Nonprobability sampling Random sampling Most basic type of probability sampling Draw numbers out of a hat/random number generator Costs less time and labor than a census A sample that looks like the population Incorporates variation Avoids bias Systematic sampling An acceptable approximation of a random sample Choose every 7th case etc. Useful when there is a complete list of elements Does not work when lists are alphabetical or numerical Stratified/proportionate sampling Useful if you want the sample to be representative of a particular population (% of women etc) Divide the sample into different categories then take % of women, % of young, % of white etc. Then you can accurately detect differences between groups Disproportionate sampling Might WANT to disproportionately sample a certain group (young black men etc) Useful when you want opinions of a particular group Reduces the number of cases needed to be representative Increases the likelihood of error Cluster sampling Divide the population into a number of small groups according to some criterion (geography eg. Hot spots policing) Select a random sample of clusters Combine all elements from the clusters into a sample Does not require as much information on the population specifics Multistage sampling Combines cluster sampling with individual element sampling Randomly select 10 districts Obtain lists of census tracts and randomly select census tracts from each of the districts Obtain list of registered voters in each of the selected census tracts Randomly select 100 to be interviewed Non-probability sampling Saves time and money Use when randomness cannot be achieved Accidental sampling (whoever happens to be at the mall that day) Purposive sampling (you select the individuals to participate based on certain criteria) Snowball sampling (convenient way to gather details when little information exists) ...
View Full Document
- representative, Systematic Sampling, population Element Sampling, Sample size Sampling, Disproportionate sampling