Chapter 4 Stats Flashcards
Population:
Wider group you study, often described as the set. Ex: IKEA survivors
Units
Each individual piece of the population. EX: Jared, Meatball Guy
Census
Collecting data on the entire sample
Bias:
Describes a method with the tendency to over/underrepresent population amounts.
Nonrepresentative:
Describes a sample that falsely represents the population. Harder to test, but biased tests are more likely to be nonrepresentative.
Sample selection bias / Sampling Bias:
Systematically too high or low sampling methods. General terms.
Size Bias:
Vague. Frequency bias, or stuff depends on physical size.
Sampling Frame
Listing all possible units in a population.
Voluntary Response Bias:
Relying on voluntary responses. Encourages passionate answers.
Convenient Sampling:
Introduces a conveience bias, where samples aren’t varied.
Judgement Sample:
Relying on an expert’s judgement to narrow down the sample. Non-probabilistic. Often quicker and more useful at the cost of experimental bias.
Non-Response Bias:
Significant portion of sample was non-responsive, effectively culling sample size.
Questionnaire Bias:
Like a push poll - “are you into the suffering of children or are you voting for George Washington?”
Incorrect Response Bias:
Lying, bad memory, possibly from a result of peer pressure.
Simple Random Sample (SRS):
All possible samples of a given fixed size are equally likely. Ex: Random number generator.
Stratified Random Sample (Strata)
Dividing units into non-overlapping subgroups, and taking a few from each. Few from all, essentially. Useful for making sure certain groups are represented.
Cluster Sample:
Dividing units into non-overlapping clusters, then selecting a few clusters and looking at every data point. All from Few.
Two Stage Sampling:
Cluster, then Strata. Dividing units into non-overlapping clusters, randomly taking a few clusters, and randomly looking at a few data points there.
Systemic Sample with Random Start
Counting off samples into non-overlapping groups, then choosing one group. Like every 8th kid in a class after starting at some random sample. Better when population size is uncertain.
Lurking Variable:
Secret third option. Obfuscated variable creating a correlation between two variables.
Treatment
Compares conditions in experimental testing. Two or more.
Response
Outcome variable in experimental testing.
Confounding:
Mixing together outcomes so effects cannot be separated, oft-accidentally. Less likely with randomizing.
Observation Study:
No treatment assigned by experimenter. Considered less reliable against confounding, as more factors may naturally confound. Factors for conditions (say, thymus size) into levels (small or big)
Clinical Trial:
Randomized & controlled experiment, often deploying double blind (researcher & patient) unawareness on placebo sample.
Experimental Units:
Groups (people, animals, families, classrooms) a treatment is assigned to. Analogous to sample size. Ex. Getting 700 cell samples for two dif sponges, meaning a sample of one for the purposes of comparison.
Replication:
Repeating treatment on different units.
Balanced:
Treatments have same number of units assigned
Blocking:
Grouping similar units in a sample and taking levels of that blocked variable to minimize confounding. Similiar to controlled variables.