Kate Storrs Module Flashcards
(16 cards)
What’s an example of a theoretical construct operationalised as a measureable variable?
Stress Perceived Stress Questionnaire (Levenstein et al., 1993) - Stress
Raven’s Progressive Matrices (Standard Ed., 1938) - Intelligence
What are main steps in NHST (Null Hypothesis Significance Test)?
State the hypothesis
Null & Alternate hypothesis formulation
Determine level of significance
Determine the test statistics
Compute the Test Statistics
Calculate P-Value
Compare the P-Value with Level of Significance
Reject or Fail to Reject Null Hypothesis
What does it mean if a test statistic is within the critical region of a sampling distribution?
Means test values are rare enough to be interesting.
For one-sided (directional) result implies results falls in 5% of data
For two sided (directional) implies results fall in 2.5% of data on one of the sides
How many types of T-Tests are there, and what are they used for?
one sample t-test - does the mean differ from null (expected value)
independent samples t-test - different individuals in each predictor category - mean difference within same individuals
paired samples t-test - same individuals in both predictor categories - mean score difference between two groups
What’s a one-way factorial ANOVA?
Statistical test used to compare the means of 3 or more groups based on one independent variable
What’s the difference between a main effect and an interaction? And how do you tell from a plot which is present?
A significant main effect tells us there is a difference between the groups - need followup pairwise comparisons to show which group means are higher or lower than another
A significant interaction occurs when one factor has a different effect depending on the level of another factor - need followup pairwise comparisons to say which factor affects the other
Follow up with TukeyHSD
Why do some tests require follow-ups?
Because they tell us that there is an interaction or main effect, however they do not tell us necessarily what that is.
If your test has low statistical power, what does that mean? And how do you fix it?
Low statistical power means there’s a high chance of missing a real effect (Type II error). It makes your test less likely to detect significant results, even if they exist.
To increase power, you can:
Increase your sample size (most effective)
Use more reliable measures
Choose a within-subjects design
Target larger effect sizes
Slightly raise the alpha level (e.g., from .01 to .05—use with caution)
More power = better chance of finding real effects.
Which pieces of information should you report from the R output for a t-test?
Mean, SD, Significance Level, T-Value, CohensD, Degrees of Freedom (df)
What is the “gist” of what a formula for variance tells us? How about the F-value?
If variance large, scores spread out, if variance is low, scores are close
F-value summarises how large between-group differences are compared to within-group differences. F-value can only be positive
Can you count how many variables are described in a research scenario, which ones are predictors vs outcomes, and whether they are continuous or categorical?
-
When do you use a two-sided vs one-sided test? A parametric vs non-parametric test?
Two-sided - when testing for difference in either direction
One-sided - when testing in a particular direction
parametric - when data meet assumptions e.g. normality, variance - t-test, anova
non-parametric - e.g. wilcoxon signed/rank-sum test, Mann-whitney U
What is the law of large numbers?
A statistic (e.g. mean) calculated from a sample approaches its true value in the whole population as the sample size grows larger
When to use correlation and/or regression?
Covariance and Pearson Correlation both measure how strong linear relationship with variables is
4 Components of Power Analysis
Sample Size
Effect Size
Power (1-beta)
Significance Level - Don’t use in power analysis test
What happens if normality assumption is not met?
One sample t-test - wilcoxon signed rank test
Paired samples t-test - wilcoxon signed rank test
Independent samples t-test - wilcoxon rank-sum test
