Research Design and Statistics Flashcards
True experimental research
At least one IV is manipulated and subjects are randomly assigned.
Quasi-experimental research
One IV manipulated, but no random assignment of subjects, typically because they’re already in groups.
Between groups design
Only compares groups that are independent. Ex: differences in reading levels of two different classes of second graders.
Within subjects design
Subjects are repeatedly measured. Ex: Looking at people’s ability to recall nouns/verbs/nonsense; all subjects are given the list of words to remember.
Analogue research
Evaluates treatment conditions that only resemble or approximate clinical situations.
Cross-sectional research
Looks at differences across sections (e.g., different ages) by sampling subjects from the age categories of interest at one point in time.
ex: How much time someone spends on the internet in groups of persons from different ages.
Longitudinal research
following a group of subjects over many years in order to understand the changes that take place as people age.
Simple random sampling
every member of the population has an equal chance of being selected.
Stratified random sampling
Population is first divided into strata (age, levels of income) and then random sample from each stratum is selected.
Threats to internal validity - History
Specific incidents that intervene between measuring points, either in or out of the experimental situation.
EX: An earthquake occurring right before implementation of a preparedness course.
Threats to internal validity - Maturation
Factors that affect the subjects’ performance because of the passing of time.
Threats to internal validity - Test practice
Familiarity with testing affects scores on repeated testing.
Threats to internal validity - Instrumentation
Changes in observers or the calibration of equpiment.
Threats to internal validity - Statistical regression
Regression to the mean; extreme scores tend to become less extreme over time.
Threats to internal validity - Selection bias
Caused by non-random assignment.
EX: if the first 20 subjects who volunteer for a study are assigned to one Tx, the next 20 to another, then we have selection bias.
Threats to internal validity - Diffusion
Occurs when the no-treatment group actually gets some of the treatment.
EX: Inadvertently having a discussion about cognitive strategies in a control group.
Construct Validity
Actually measuring what you intend to.
Threats to construct validity - Attention & Contact with Clients
Hard to tell whether the change in clients is due to the intervention or the attention/contact.
Threats to construct validity - Experimenter Expectancies
Cues/clues transmitted to the subjects by the experimenter ultimately affecting the data.
Rosenthal effect.
Threats to construct validity - Demand characteristics
Factors in the procedures that suggest how a subject should behave.
Threats to construct validity - John Henry Effect
Compensatory rivalry. Persons in the control group try harder than usual in the spirit of competition.
Threats to external validity - Contextual characteristics
Conditions in which the intervention is imbedded. Reactivity occurs when subjects behave in a certain way just because they are participating in research and being observed.
Standard deviation
Average deviation from the mean in a given set of scores. Square root of the variance.
Sampling error
samples drawn from populations are usually not perfectly representative of the population.
Central limit thoerem
Assuming an infinite number of equal sized samples are drawn from a population, the means will be normally distributed.
Type I error
rejecting the null when it’s really a mistake; significance is found but subsequent research finds no significance.
most likely in t-tests
Type II error
Accepting the null when you should reject it; significance is not found in original EXP, but subsequent studies find significance.
Power
Defined as ability to correctly reject the null.
Coefficient of determination
Squaring the correlation coefficient. Represents the amount of variability in Y that is shared with, xplained by, or accounted for by X.
EX: r = .5; coeff of det: .25. 5% of variability in Y is explained by X.
Assumptions of Bivariate correlations (3)
- Linear relationships
- Homoscedasticity: similar spread of scores across the entire scatter plot.
- Unrestricted range
Zero-order correlation
Relationship between x and y when it is believed that no extraneous variables affect the r/s.
Partial correlation
Looking at r/s b/w two variables while controlling for effect of a third.
Part correlation
Examines r/s between predictor and criterion w/ the influence of the third variable removed ONLY for one of the variables.
Multivariate tests
Tests of correlation and prediction involving several IVs and one or more DVs
Multiple R
Correlation between two or more IVs and one DV, where the DV is always interval or ratio.
Multiple regression
Allows the prediction of the DV based on the values of the IVs.
Multicollinearity
Occurs in a multiple regression equation when the predictors are highly correlated with one another.
Discriminant function analysis
Same as multiple regression, but the DV is nominal (categorical; pass/fail)
Path analysis
Applies multiple regression to test a model that specifies causal links between variables.
Factor analysis
Operates by extracting as many significant factors from the data as possible.
Eigenvalues
Indicator of strength of a factor.
Test retest reliability
Coefficient of staiblity; consistency in scores on same measure at different time points.
Parallel forms of reliability
Coefficient of equivalence; Correlating scores obtained by the same group on two equivalent but not identical forms of the same test.
Internal consistency
Consistency of scores within a test.
Standard error of measurement
Standard deviation of a normal distribution of scores obtained by one individual on equivalent tests.
Content validity
How adequately a test samples a particular content area
Criterion related validity
How adequately a test score can be used to infer, predict, or estimate an outcome (e.g., how well does GPA predict SAT scores).
Concurrent validity
Predictor and criterion are measured and correlated at the same time.
Predictive validity
Using one score to predict another later on.
Standard error of estimate
Average amount of error in estimating an outcome based on a predictor.
Construct validity
How adequately a new test measures a construct or trait. Assessed with a factor analysis.
Convergent validity
Correlation of scores on the new test with other available measures.
Discriminant validity
Correlation of scores on the new test with scores on another test that meaures a different trait. Low discriminant validity between two diff measures means high construct validity.
If a constant is subtracted from every score in a sample, what happens to mean/SD?
Mean: Decrease
SD: same
Pooled error term is used in ANOVA when?
Sample size is unequal.
Biggest issue with single subject research design is
Autocorrelation
Cluster sampling
Researchers divide a population into smaller groups.
Structural equation modeling
Multiple pathways involving multiple predictors and multiple criterion variables.
Face validity
whether something seems like it’s measuring what it’s supposed to based on face value
When to use ANCOVA
When unexpected differences are uncovered among treatment groups with regard to an extraneous variable.
What is likely to happen when a kid who tested very high on the WISC takes it again in 3 years.
Lower.
LISREL, a form of SEM, can be used to
test a causal model of relationships among variables
To improve reliability on a new test you’re developing, what should you do?
Use a heterogeneous sample.
How should you analyze your results if you’re looking at whether several variables can predict one nominal (pass/fail) outcome?
Discriminant analysis
Sequential introduction of treatment is closest to what research design?
Multiple baseline.
In normal distribution, bell curve, what is the percentile of +/- 1 SD from the mean.
+1: 84% and up
-1: 15.9% and down.
+2: 98%
-2: 2%
How to increase chance of finding significance?
Increase sample size, use a one-tailed test, increase alpha.
Kappa coefficient
Interrater reliability
Split plot anova
used to test for differences between two or more independent groups whilst subjecting participants to repeated measures.
Standard errors of the mean, measurement, and estimate express error in terms of _______
the Standard Deviation
Relationships of the standard error of estimate
Direct relationship with the SD of the criterion.
Indirect relationship with validity.
Significant differences in a 2-way ANOVA means what
Any combination of main effects and interactions.
Relationship between education and income among clinical psychologists?
0.0