510 likes | 624 Views
Chapter 2. Methods and Statistics in I-O Psychology. Royalty-Free/CORBIS. Module 1: Science. What is science? Science has common methods Science is a logical approach to investigation Based on a theory, hypothesis or basic interest Science depends on data
E N D
Chapter 2 Methods and Statistics in I-O Psychology Royalty-Free/CORBIS
Module 1: Science • What is science? • Science has common methods • Science is a logical approach to investigation • Based on a theory, hypothesis or basic interest • Science depends on data • Gathered in a laboratory or the field
Common Methods (cont'd) • Research must be communicable, open, & public • Research published in journals, reports, or books 1) Methods of data collection described 2) Data reported 3) Analyses displayed for examination 4) Conclusions presented
Common Methods (cont'd) • Scientists set out to disprove theories or hypotheses • Goal: Eliminate all plausible explanations except one • Scientists are objective • Expectation that researchers will be objective & not influenced by biases or prejudices
Role of Science in Society • Expert witnesses in a lawsuit • Permitted to voice opinions about practices • Often a role assumed by I-O psychologists
Daubert Challenge • Challenging testimony of an expert on the grounds it is not scientifically credible • Daubert v. Merrill-Dow, 1993 • Resulted in introduction of a method for distinguishing between “legitimate science” & “junk science”
Scientific Testimony in Court • Theories presented in court must: • Be recognized by particular scientific area as worthy of attention • Be peer reviewed or subjected to scientific scrutiny • Have a known “error rate” • Be replicable or testable by other scientists
Module 1 (cont'd) • Why do I-O psychologists engage in research? • Better equip HR professionals in making decisions in organizations • Provide an aspect of predictability to HR decisions
Module 2: Research • Research design • Experimental • Random assignment of participants to conditions • Conducted in a laboratory or the field • Quasi-experimental • Non-random assignment of participants to conditions
Research Design (cont'd) • Non-Experimental • Doesn’t include “treatment” or assignment to different conditions • 2 common designs: • Observational design • Survey design
Methods of Data Collection • Quantitative methods • Rely on tests, rating scales, questionnaires, & physiological measures • Yield results in terms of numbers C. Borland/PhotoLink/Getty Images
Methods of Data Collection • Qualitative methods • Include procedures like observation, interview, case study, & analysis of written documents • Generally produce flow diagrams & narrative descriptions of events/processes
Quantitative & Qualitative Research • Not mutually exclusive • Triangulation, (Rogelberg & Brooks, 2002) • Examining converging information from different sources (qualitative and quantitative research).
K Lewin • B = f (p*e) • Behavior is a function of • Person X environmental influences
Experimental v. Corr research • “I” side: focus on Individual differences • Person attributes: • E.g. Personality, behaviors, cognitive ability • “O” side: focus on Environmental influences • Situation variables: • E.g. work conditions, leadership style, pay for performance
I v. O • Which is most likely to use • Experimental designs? • Correlational designs? • Why?
Generalizability in Research • Application of results from one study or sample to other participants or situations • Benefit of using theory • Every time a compromise is made, the generalizability of results is reduced
Sampling Domains for I-O Research Figure 2.1: Sampling Domains for I-O Research
Observational Unit • Worker • Team • Department • Organization • Industry • Others?
Measurement Unit(one of something) • Identify a measurement unit for: • Worker’s performance score • Years of experience • Absenteeism • Motivation • Sales performance • Cognitive ability
Control in Research • Experimental control • Influences that make results less reliable or harder to interpret are eliminated • Statistical control • Statistical techniques used to control for influences of certain variables
Ethics • Ethical standards of the APA • Collection of 61 cases endorsed by SIOP • Illustrates ethical issues likely to arise in I-O psychology (Lowman, 1985a,1998)
Module 3: Data Analysis • Descriptive statistics • Summarize, organize, describe sample of data Frequency Distribution: • Horizontal axis = Scores running low to high • Vertical axis = Indicates frequency of occurrence
Describing a Score Distribution • Measures of central tendency • Mean • Mode • Median Ryan McVay/Getty Images
Describing Score Distribution (cont'd) • Variability • Standard deviation • Lopsidedness or skew Ryan McVay/Getty Images
Descriptive Statistics:Two Score Distributions (N = 30) Figure 2.2 Two Score Distribution (N=30)
Two Score Distributions (N = 10) Figure 2.3
Inferential Statistics • Aid in testing hypotheses & making inferences from sample data to a larger sample/population • Include t-test, F-test, chi-square test
Statistical Significance • Defined in terms of a probability statement • Threshold for significance is often set at .05 or lower • p < .05 (likelihood of this effect size would occur less than 5 times in a hundred)
Statistical Power • Likelihood of finding statistically significant difference when true difference exists • Smaller the sample size, lower the power to detect a true or real difference
Concept of Correlation Positive Linear Correlation Figure 2.4 Correlation between Test Scores and Training Grades
Concept of Correlation (cont'd) • Scatterplot • Displays correlational relationship between 2 variables • Regression • Straight line that best fits the scatterplot
Correlation Coefficient • Statistic or measure of association • Reflects magnitude (numerical value) & direction (+ or –) of relationship between 2 variables
Correlation Coefficient • Positive correlation → High values of one variable are associated with high values in the other variable (& vice versa) • Negative correlation → High values of one variable are associated with low values in the other (& vice versa)
Curvilinear Relationship • Although correlation coefficient might be .00, it can’t be concluded that there is no association between variables • A curvilinear relationship might better describe the association (eta η) • SPSS can provide η with F test
Curvilinear Correlation Figure 2.7 An Example of a Curvilinear Relationship
Multiple Correlation • Multiple correlation coefficient • Overall linear association between several variables & a single outcome variable (R) • R2 = Proportion of variance in DV (outcome) accounted for all preditors (several vars)
Meta-Analysis • Statistical method for combining results from many studies to draw a general conclusion • Statistical artifacts • Characteristics of a particular study that distort the results • Sample size is most influential
Module 4: Interpretation • Reliability • Consistency or stability of a measure • Test-retest reliability • Calculated by correlating measurements taken at Time 1 with measurements taken at Time 2
High and LowTest-Retest Reliability Figure 2.8 Examples of High and Low Test-Retest Reliability: Score Distributions of Individuals Tested on Two Different Occasions
Reliability (cont'd) • Equivalent forms reliability • Calculated by correlating measurements from a sample of individuals who complete 2 different forms of same test • Internal consistency (Cronbach alpha α) • Assesses how consistently items of a test measure a single construct
Reliability (cont'd) • Inter-rater reliability • Can calculate various statistical indices to show level of agreement among raters • Intraclass correlation ICC • Rwg • Generalizability theory • Simultaneously considers all types of error in reliability estimates
Validity • Whether measurements taken accurately & completely represent what is to be measured • Predictor • Test chosen or developed to assess identified abilities • Criterion • Outcome variable describing important aspects or demands of the job
Figure 2.9: Validation Process from Conceptual and Operational Levels Figure 2.9
Criterion-Related Validity • Correlate a test score with a performance measure (validity coefficient) • Predictive validity design • Time lag between collection of test data & criterion data • Test often administered to job applicants
Criterion-Related Validity (cont'd) • Concurrent validity design • No time lag between collection of test data & criterion data • Test administered to current employees, performance measures collected at same time • Disadvantage: No data about those not employed by the organization
Content-Related Validity • Demonstrates that content of selection procedure represents adequate sample of important work behaviors & activities or worker KSAOs defined by job analysis
Construct-Related Validity • Investigators gather evidence to support decisions or inferences about psychological constructs • Construct - concept or characteristic that a predictor is intended to measure; examples include intelligence and extraversion
A Model for Construct Validity Figure 2.10 A Model for Construct Validity