Term
|
Definition
| When the probability of selecting each sampling unit is unknown |
|
|
Term
| Simple Random sampling(prob) |
|
Definition
•A sampling approach in which each sampling unit in a target population has a known and equal probability of being included
• Advantage: Good generalizability and unbiased estimates
•Disadvantage: must be able to identify all sampling units within
a given population; often, this is not feasible |
|
|
Term
| Systematic random sampling(prob) |
|
Definition
•Similar to random sampling, but work with a list of sampling units that is ordered in some way (e.g., alphabetically).
•Select a starting point at random, then survey each nth person where the “skip interval” = (population size/desired sample size)
•Advantage: quicker and easier than SRS
•Disadvantage: may be hidden “patterns” in the data |
|
|
Term
| Stratified random sampling(prob) |
|
Definition
•Break up population into meaningful groups (e.g., men, women),
then sample within each “strata”, then combine
•Proportionate stratified sampling: here you sample based on the size of the populations (i.e., sample more from the bigger strata: e.g., Caucasians)
•Disproportionate stratified sampling: sample the same number of units from each strata, regardless of the strata’s size in the pop.
•A variant is optimal allocation: here you use smaller sample sizes for strata within which there is low variability (as the lower variability will give you more precision with lower N).
•Advantages: more representative; can compare strata
•Disadvantages: Can be hard to figure out what to base strata on (Gender? Ethnicity? Political party?) |
|
|
Term
|
Definition
-Similar to stratified random sampling, but with stratified random sampling, the strata are thought to possibly differ between strata (men vs. women), but be homogeneous within strata.
-In cluster sampling, you divide overall population into subpopulations (like SRS), but each of those subpopulations (called “clusters”) are assumed to be mini-representations of the population (e.g., survey customers at 10 Red Robins in WA).
-Area sampling: clusters based on geographic region |
|
|
Term
| Cluster sampling steps (prob) |
|
Definition
•One-step clustering: just select one cluster (e.g., one store);
problem = may not be representative of population
•Two-step cluster sampling: break into meaningful subgroups (Red Robins in big cities vs. Red Robins in suburbs), then randomly sample within each of those clusters
•Advantages: easy to generate sampling frame; cost efficient; representative; can compare clusters
•Disadvantages: must be careful in selecting the basis for clusters; also, within clusters, often little variability (they’re homogeneous), and this lack of variability leads to less precise estimates |
|
|
Term
| Convenience sample non prob |
|
Definition
•Survey people based on convenience (e.g., college students)
•Advantage: is fast and easy
•Disadvantage: may not be representative |
|
|
Term
| Judgement sampling non prob |
|
Definition
•Use your judgment about who is best to survey
•Advantage: Can be better than convenience if judgment is right
•Disadvantage: but if judgment wrong, may not be representative/generalizable |
|
|
Term
|
Definition
•Sample fixed number of people from each of X categories, possibly
based on their relative prevalence in the population
•Advantage: Can ensure that certain groups are included
•Disadvantage: but b/c you aren’t using random sampling, generalizability may be questionable |
|
|
Term
| Snowball sampling non prob |
|
Definition
•You contact one person, they contact a friend (e.g., one cancer
survivor is in contact with other survivors, and so recruits them)
•Advantages: can make it easier to contact people in hard to reach groups
•Disadvantage: there may be bias in the way people recruit others |
|
|
Term
| Factors affecting choice of sampling procedure |
|
Definition
•You are collecting quantitative data that you want to use to arrive at
accurate generalizations about population
•You have sufficient resources and time •You have a good sense for the population
•You are sampling over a broader range (e.g., of states, nations) |
|
|
Term
|
Definition
| Any type of bias that is attributable to mistakes in either drawing a sample or determining the sample size |
|
|
Term
|
Definition
•A bias that occurs in a reesearch study regardless of whether a sample or a census is used (recall all the different types of errors we discussed)
•Respondent Errors (non response, response errors)
• Researcher’s measurement/design errors (survey, data analysis)
• Problem definition errors
• Administrative errors (data input errors, interview errors, poor sample design) |
|
|
Term
|
Definition
| The difference is that statistics describe a sample, whereas a parameter describes an entire population. |
|
|
Term
|
Definition
•Identifying and defining what is to be measured
•A construct is a hypothetical variable composed of different elements that are thought to be related (e.g., 5 questions tapping brand loyalty) |
|
|
Term
|
Definition
•Figuring out how to measure what you want to measure
•Measure needs to be reliable and valid |
|
|
Term
|
Definition
•Extent to which items on a scale “hang together” or are correlated with one another
• Cronbach’s alpha (covered in last class)
• Split-half reliability (split measure into halves, correlate) |
|
|
Term
|
Definition
•Extent to which scores are stable over time
•Have people complete questionnaire twice and correlate scores |
|
|
Term
|
Definition
| The extent to which conclusions drawn from a study are true |
|
|
Term
|
Definition
| When a researcher can clearly identify cause and effect relationships (i.e., there are no confounds) |
|
|
Term
|
Definition
| The extent to which what you find in your study can be generalized to your target population |
|
|
Term
|
Definition
•Extent to which your constructs of interest (e.g., sensation seeking) are accurately and completely identified (measured)
•In other words, the extent to which you are actually measuring what you say you are measuring (your sensation seeking scale really does measure the true construct of sensation seeking) |
|
|
Term
|
Definition
Extent to which a measure is appropriate according to experts in the
domain of interest |
|
|
Term
| Concurrent validity (CONVERGENT) |
|
Definition
Extent to which one measure of a construct overlaps with other
similar measures of that construct |
|
|
Term
|
Definition
Extent to which a measure of one construct does not overlap with
measures of different constructs |
|
|
Term
|
Definition
Extent to which a measure of a construct can predict theoretically-
relevant outcomes |
|
|
Term
|
Definition
| How a construct fits within a broader set of related constructs |
|
|
Term
|
Definition
assignment: you can assign objects to categories
order (magnitude): you can order objects in terms of having more or less of some quality
distance-equal intervals: the ditance between adjacent points on the scale is identical
Origin-absolute zero: zero "means something" (absence of a given qaulity) |
|
|
Term
|
Definition
Nominal- has assignment only (political party)
ordinal: has assignment, order, (rank order of finish in a race)
interval: has assignment, order, equal intervals (temp)
hybrid ordinally-interval scale:Like an ordinal scale, but researcher “pretends” it is an interval scale (e.g.,
assumes 1 to 7 scale is an interval scale); commonly used in questionnaires
Ratio: Has Assignment, Order, Equal Intervals, Absolute Zero (Number of Cars) |
|
|
