Chapter 12: Understanding Research Results

Chapter 12:Understanding ResearchResultsPsychology 2301Introduction to Research MethodsUniversity of Houston

Scales of Measurement Remember these?– Nominal– Ordinal– Interval– Ratio

Describing the Results How you describe the results of a studydepends on the measurement scales ofthe variables. There are three basic types of description:– Comparing group percentages– Correlating scores on two variables– Comparing group means

Comparing Group Means Used when IV is nominal (or possibleordinal) and DV is interval or ratio. Example: Middlemist, Knowles, & Matter(1976).

Comparing Group Percentages Used when both the IV and DV are on anominal (or possibly ordinal) scale Example: Buss, Larsen, Westen, &Semmelroth (1992)

Correlating Individual Scores Used when:– There are no distinct groups of participants– Both the predictor and criterion are measuredon an interval or ratio scale (See Pearson r)

Frequency Distributions An arrangement of a set of scores fromlowest to highest that indicates the numberof times each score was obtained Graphs– Pie chart (Figure 12.1) Nominal data– Bar graphs (Figure 12.2) Nominal data– Frequency polygons (Figure 12.3) Interval/Ratio data

Descriptive Statistics Statistical indices that describe data Types– Central tendency– Variability– Correlation

Central Tendency: Mean Average Abbreviated as M and X (“x bar”) Used with interval or ratio data– Why not with nominal? Sensitive to extreme values

Central Tendency: Median The middle value of a set of scoresAbbreviated as MdnUsed with ordinal, interval, and ratio dataNot sensitive to extreme scores

Central Tendency: Mode The most frequent score Only measure of central tendency that isuseful with nominal data

Variability The amount of spread in a distribution ofscores Two types– Standard deviation– Variance(s2)

Variability Standard Deviation– Abbreviated as s or SD– Standard deviation is small when scores areclose to the mean– Used with interval and ratio data Variance– Abbreviated as s2

Correlation Describes the relation between two or morevariables There are multiple types of correlation, but thePearson r is the most common.– Comprised of magnitude and direction– Values range from – 1 to 1 - 1: Perfect negative correlation 1: Perfect positive correlation 0: No relation (horizontal line)– The closer to 1/-1 the stronger the relation betweenvariables– Which is stronger? -.67 or .23– Figures 12.6 & 12.7 (pp. 222-3)

Pearson r: Other Considerations Restriction of range– Reduces the magnitude of the correlationbetween variables Curvilinear relation– Results in no relation between variables

Effect Size Measure of the strength of associationbetween two variables. Pearson r is one measure of effect size,but effect sizes can be calculated for alldifferent types of variables. r - ranges from 0.00 to 1.00 (no direction)– small (.15), medium (.30), large (.40) r2 - range from 0% to 100% (of sharedvariance)

Regression A mathematical equation that allows forthe prediction of a score on one variablewhen the score on another is known. Y bX a Example: Depression predicting SuicideAttempt.

Multiple Correlation (R) An index of the relation between severalpredictor variables and a single criterionvariable R2 (similar to r2) is a measure of sharedvariance between the criterion and the setof predictors.

Multiple Regression Regression that encompasses more thanone predictor variable Y b1X1 b2X2 bnXn a Example: Depression and Genderpredicting Suicide Attempt.

Partial Correlation Controls for third variables Assess the relation between variables ofinterest after the influence of the thirdvariable is removed Figure 12.9 (pg. 229)

Structural Models Hypothesized relation among variablesAttempt to “fit” data to modelArrows indicate relation among variablesFigure 12.10 (pg. 212)

Regression vs. SEMCABD

Chapter 12: Understanding Research . Results. Psychology 2301. Introduction to Research Methods. . There are multiple types of correlation, but the Pearson r is the most common. – Comprised of magnitude and direction . Regr