Research Design/Stats Flashcards
Threats to Internal Validity (things other than IV that affect outcome)
§ History
§ Maturation
§ Testing or test practice
§ Instrumentation (user errors)
§ Statistical Regression (regression to the mean- extreme scorers move closer to the mean)
§ Selection Bias (nonrandom assignment)(quasi experiments)
§ Attrition or experimental mortality
§ Diffusion (treatment affects next treatment-carryover)
Factors affecting research outcomes
Demand Characteristics (subjects)
§ Aspects that suggest how subjects should bx
§ Controlled by seeping subjects blind to their condition
o Expectancy (experimenters)
§ Experimenter cues-Rosenthal Effect (late bloomers)
§ Experimenter inadvertently giving cues about how to bx
§ Controlled by blinding experimenter
o Reactivity
§ Subject’s responses that occur simply as a result of participating in research (being observed)
§ Hawthorne Effect- productivity increased just bc they were being observed
Threats to external Validity (generalizability)
§ Sample characteristics (representative?)(random selection controls for this issue)
§ Stimulus characteristics (like the real-world?)
§ Contextual characteristics (like the real-world?)
Mediator Variable
explains the relationship
Moderator Variable
affects the intensity of the relationship
Type I Error
·A Type I error means rejecting the null hypothesis when it’s actually true. It means concluding that results are statistically significant when, in reality, they came about purely by chance or because of unrelated factors. The risk of committing this error is the significance level (alpha or α) you choose
False Positive
Type II Error
Failure to Reject the null when false. False negative
Beta-prob of making a type II error
Positive Skewed
- mode, med, mean
Neg Skewed
Mean, Med, mode
Standard error of the mean (SEM)
is the std dev divided by the square root of the sample size (N), as std dev increases so does std error of the mean, the larger the sample the less std of error (less prone to error and more representative)
Parametric tests
Parametric tests of difference (there are parameters that must be met-normally disturbed data, interval ratio data (score value), homocedasity (equal spread or variability, equal std)
ANOVAs
One-way ANova- one IV, Two-way if two IV, three-way- three IV (two or more groups)
o Factorial Anova- two-ways or more (two-way Anova or more) but all IV data has to be independent (no related data) (time cannot be analyzed bc they are correlated)(no repeated measures)
o Split-plot Anova (Mixed anova)-two IVs, one IV is independent, one IV is correlated (i.e. time)
ANCOVA
ANCOVA
o Combines Anova and partial correlation (partials out the effects of a moderator)(controls for a third or extraneous variable)
Regressions/Predictors
o Multiple Regression Equation- (regression is a prediction- Simple regressions one variable predicts another) Multiple-multiple predictors of an outcome (predictors are compensatorythey can compensate for one another (i.e. fewer calories and exercising help with weight loss))
o Multiple Cutoff- non-compensatory-applicants must meet or exceed the cutoff on each predictor (i.e. high IQ and dexterity scores to become surgeons)(conjunctive)
o Multiple Hurdle- non-compensatory, predictors applied in a certain order, must pass cutoff on one test to move on to the next (i.e. submit an application before can be considered for interview)
Bivariable correlation tests
o Dichotomous variables (either/or)- PHI (rarely possible- both have to be true dich)
o Continuous variables- Pearson R (both are continuous)
o Spearmen’s Rho/ Kendall TAU both variables are ordinal (rank ordered)
o ETA-continuous variables with curvilinear relationship (extremes are low or high-optimal stress)
o Biserial if one is continuous and the other is an artificial dichotomy (i.e. college education or not)
o To calculate shared variability between two correlations just square the correlation coefficient of each and then compare (ie. .9 and .3 becomes .81 compared to .9, the shared variability in the first is 9 times greater than the second)
