Psychological Research Methods: NHST, Regression, Mediation, and Moderation

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/34

flashcard set

Earn XP

Description and Tags

Comprehensive practice flashcards defining foundational concepts in null hypothesis significance testing, effect sizes, statistical errors, regression models, mediation, and moderation.

35 Terms

1
New cards

Null Hypothesis Significance Testing (NHST)

A four-step inferential framework consisting of specifying null and alternative hypotheses, selecting a significance threshold (α\alpha), calculating test statistics with their associated pp-values, and determining whether to reject or fail to reject the null hypothesis.

2
New cards

pp-value

The probability of obtaining sample results at least as extreme as the observed data, assuming that the null hypothesis is true.

3
New cards

Effect Size

A standardized metric that quantifies the magnitude or strength of an association or experimental effect, which is not directly dependent on sample size.

4
New cards

Cohen's dd

A standardized effect size metric for assessing mean group differences (categorical IV and continuous DV), with benchmarks of 0.20.2 for small, 0.50.5 for medium, and 0.80.8 for large effects.

5
New cards

Pearson's rr

A standardized correlation coefficient and effect size metric evaluating the linear relationship between two continuous variables, ranging from −1-1 to +1+1, with benchmarks of 0.10.1 for small, 0.30.3 for medium, and 0.50.5 for large associations.

6
New cards

Meta-analysis

A quantitative research method that synthesizes standardized effect sizes across all available published studies in the literature to estimate an overall effect size.

7
New cards

Type I Error (α\alpha)

A false positive error occurring when a researcher rejects the null hypothesis when it is actually true in reality, with the decision criterion typically set at α=.05\alpha = .05.

8
New cards

Type II Error (β\beta)

A false negative error occurring when a researcher fails to reject the null hypothesis when an actual effect or relationship exists in reality.

9
New cards

Statistical Power (1−β1 - \beta)

The probability of detecting a statistically significant effect when that effect truly exists in reality, which increases as the criterion α\alpha, sample size, and effect size increase (typically targeted at .80.80).

10
New cards

Nominal Variable

A categorical level of measurement that sorts observations into distinct, non-ordered qualitative classifications.

11
New cards

Ordinal Variable

A categorical level of measurement with at least three distinct categories that possess an inherent order or ranking in magnitude, but lack equal intervals between adjacent points.

12
New cards

Interval Variable

A continuous level of measurement characterized by equal distances between adjacent scale points across the scale, but lacking an absolute, true zero.

13
New cards

Ratio Variable

A continuous level of measurement that features equal intervals between scale points and an absolute, true zero that signifies the complete absence of the property being measured.

14
New cards

Likert Scale

A rating format that is ordinal in theory, but is practically treated as an interval variable in psychological research when it contains more than three response options.

15
New cards

Replication Crisis

A methodological crisis in social and psychological sciences underscored by an Open Science Collaboration project where only 36%36\% of 100 examined studies successfully replicated.

16
New cards

File-Drawer Problem

A systemic publication bias where researchers and journals selectively publish statistically significant results while non-significant, null findings remain unwritten or locked away in file drawers.

17
New cards

pp-hacking

The misuse of data analysis techniques—such as testing multiple comparisons without corrections, stopping data collection once p<.05p < .05, or selectively excluding participants and covariates—to artificially generate statistically significant results.

18
New cards

Pre-registration

The open-science practice of officially documenting and publishing hypotheses, study design, stopping rules, and statistical analyses prior to data collection to prevent pp-hacking.

19
New cards

zz-score

A standardized score indicating the exact number of standard deviations a raw score lies above or below the mean, computed as z=x−xˉsdz = \frac{x - \bar{x}}{sd}.

20
New cards

Centering

The mathematical transformation of subtracting the mean from each raw score (C=x−xˉC = x - \bar{x}), shifting the mean to zero while preserving the original measurement scale units.

21
New cards

Unstandardized Regression Coefficient (bb)

The regression slope parameter representing the predicted change in the dependent variable in raw units for every 1-unit increase in the independent variable.

22
New cards

Standardized Regression Coefficient (β\beta)

The regression slope parameter representing the predicted change in the dependent variable in standard deviation units for every 1-standard deviation increase in the independent variable.

23
New cards

Regression Intercept

The predicted value of the dependent variable when all independent variables in the regression model are equal to zero.

24
New cards

Cronbach's Alpha (α\alpha)

A measure of internal consistency reliability ranging from 00 to 11, calculated from the number of items and their average inter-item correlation, where values between .65.65 and .79.79 are acceptable and values above .80.80 are good.

25
New cards

Multiple Regression

A statistical analysis that predicts a single continuous dependent variable using multiple independent variables simultaneously, determining each variable's unique contribution while controlling for the others.

26
New cards

Partial Correlation

A correlation quantifying the unique linear association between two variables after statistically removing the shared variance of a third confounding variable from both.

27
New cards

Prospective Change Analysis

A longitudinal regression technique that predicts a dependent variable at Time 2 while controlling for its baseline level at Time 1, isolating true changes over time from baseline stability.

28
New cards

Mediator

An intervening variable that explains the underlying causal mechanism or pathway through which an independent variable influences a dependent variable, addressing 'how' or 'why' an effect occurs.

29
New cards

Moderator

A variable that alters, qualifies, or changes the direction or strength of the relationship between an independent variable and a dependent variable, addressing 'when' or 'for whom' an effect occurs.

30
New cards

Indirect Effect

The specific mediated pathway linking the independent variable to the dependent variable through a mediator, quantified statistically as the product of the regression coefficients a⋅ba \cdot b.

31
New cards

Sobel Test

A classical significance test used in mediation analysis to assess whether the reduction in the direct effect of an IV on a DV after adding the mediator is statistically significant.

32
New cards

Bootstrapping (in Mediation)

A non-parametric resampling procedure with replacement that creates an empirical sampling distribution to calculate confidence intervals and test the significance of the indirect effect (a⋅ba \cdot b).

33
New cards

Main Effect

The overall effect of a single predictor variable on the dependent variable, averaged across all levels of the other predictor or moderator variables in the model.

34
New cards

Interaction Effect

A statistical effect occurring when the relationship between an independent variable and a dependent variable differs depending on the level of a moderator variable, reflected visually by non-parallel lines.

35
New cards

Simple Slopes Analysis

A statistical decomposition technique conducted after detecting a significant interaction to calculate and test the significance of the IV-DV slope at specific representative levels of the moderator (typically −1 SD-1\text{ SD}, Mean, and +1 SD+1\text{ SD}).