# PSYC 2030 Study Guide - Midterm Guide: Intraclass Correlation, Correlation And Dependence, Regression Analysis

by OC382513

School

York UniversityDepartment

PsychologyCourse Code

PSYC 2030Professor

Rebecca JubisStudy Guide

MidtermThis

**preview**shows page 1. to view the full**4 pages of the document.**Factorial design: any study with >1 IV

•Ex: 2 x 3 x 4 = 3 IVs, 2 levels on 1st, 3 lvls on 2nd, and 4 levels on 3rd

•Factorial matrix:

oLevels: apply to a single IVs

oConditions: all the possible IV variations added together (ex: 2x3x4)

•Main effect: effect of a single IV, compare all levels of one factor vs another

oCalculate row /column means

•Interactions: when effect of one IV depends on level of another IV

oCan occur without a main effect present

oCan trump main effect so that the main effects are actually misleading if reported

as is.

Hint: when 1 condition varies far from all others, with all others being

similar)

oOn graph: when lines are not parallel, much easier to see than bar graphs, thus

recommended over bar graphs when interaction is suspected (even if IV is

discrete)

•mixed design: at least 1 b/w subject variable, and 1 w/i subjects

oCounterbalancing required for the w/i variable

oNot required if examining “trials” in learning studies, which actually looks at

changes with trials thus have to be in same order.

•P x E: both manipulated and subject variables

oCan draw causality if main effect occurs only for E but not P and no interaction

oATI designs (aptitude treatment interaction)

•Mixed PxE: if the manipulated variable is a w/i variable (thus study has b/w and w/i, as

well as subject and manipulated), the subject variable MUST be a b/w

•ANOVA:

o1 factor = 1F; 2 factor = 3Fs (2 + interaction); 3 factor = 7 Fs

oSimple effects analysis: comparing each levels of one factor to each levels of

another,

Correlational design

•Positive: both variables change in same direction

•Negative: inverse relationship

•Pearson’s r (coefficient of correlation) for interval/ratio scale,

or^2: portion of variability in one variable that can be accounted for by variability in

the other.

•Spearman’s rho for ordinal data (rankings)

•Scatterplots: would fail to see complex “curved line” relationships

•Restriction of range: weakens correlation because only included data within a certain

threshold

•Regression analysis: making predictions based on correlation

oRegression line / best fitting line: distance between each point and the line are at

minimum

Y = a + bX, where X is the predictor variable and Y is the criterion (to be

calculated by a given value of X)

oShould only be used for subjects within range of values that the correlation is

based on

•Does NOT hold constants thus no conclusions

###### You're Reading a Preview

Unlock to view full version