Statistics and Research Methods Flashcards
Campbell’s validity typology
Internal validity, external validity, statistical conclusion validity, construct validity
Internal validity
Extent to which association between variables is causal in nature. Valid causal inference requires 1) statistical association, 2) temporal precedance and 3) nonspuriouslness
Case-control designs
Compares a group of participants with a certain characteristic (case group) to a group without that trait (control group).
Cohort designs
Intact group is followed over time to examine changes in outcome of interest. Design is longitudinal or prospective.
Cross-sequential design
When multiple cohorts differ in age or some other salient developmental marker at the study’s inception
Quasi-Experimental Studies
Experiments without random assignment. Subset of units is typically exposed to an intervention. Experimental designs include interrupted time series design and regression discontinuity design.
Single-Case Experiments
Common features include intensive assessment before, during, and after intervention. Prolonged baseline assessment provides information about the pattern of changes in the outcome in the absence of the intervention.
ABAB Designs
single-case design that alternates the baseline A phase (intervention absent) with an intervention B phase (intervention present). The outcome of interest is assessed on multiple occasions within each phase.
Multiple baseline designs
Replication of an effect is sought over multiple baselines, which can reflect different behaviors, settings, and/or children (just to name a few).
Hypothetical counterfactual
In RTC, what would have happened to those not in treatment condition. What you learn from control condition.
Efficacy trials
Intervention’s effects are examined under ideal circumstances, particularly with respect to treatment implementation
Effectiveness trials
Intervention’s effects are examined under real-world conditions.
Intent-to-treat analyses
Designed to analyze outcome data from randomized experiments involving participant attrition. Researchers analyze outcome data from participants as a function of their original group assignment, regardless of their level of exposure to treatment.
Levels of evidence
Mudford, McNeill, Walton, and Phillips (2012). As knowledge in an area accumulates, discussions of ordering evidence sources along a continuum from low to high become more prominent.
Threat to construct validity
Therapeutic attention (i.e. common factors), confounding clinicians with treatment, dynamics of a group for group therapy trials.
Threats to internal validity
Maturation, history, statistical regression, attrition, testing effect, instrumentation, selection and interactions with selection.
Parametric statistics
Assumption of normal distribution, data measured on interval or ratio scale
Type I error
true null hypothesis is incorrectly rejected (i.e., results are declared statistically significant even though the null hypothesis is true).
Type II error
False null hypothesis is not rejected (i.e., results are not declared statistically significant even though the null hypothesis is false).
NNT (Number needed to treat)
Measure of clinical significance. For binary outcome, reflects rate of success in treatment versus control group. When NNT is larger, effect is smaller.
One sample t-test
Used to test the difference between a single sample mean and an hypothesized population mean. Used instead of the one-sample z test if the population SD is unknown (a more typical scenario).
Student’s t-test
Independent samples t-test
Paired (related) samples t-test
each “participant” contributes a pair of data points to the analysis, which are assumed to be dependent (i.e., correlated).
Omnibus tests of significance
In models for which there are three or more levels of a factor (e.g., low, medium, and high levels of stress), the test of the factor’s main effect. Doesn’t show whether 2 levels are significantly different
Manova
More than 1 dependent variable