Data AnalysisQuality of DataValidityThe extent to which the data collected measure the intended variableErrorEffects due to variables other than those measuredNOT just noise or mistakes; includes e.g. variation among individuals2 types: Random and SystematicRandom ErrorAmount is different on each measurementTends to average out to zero over time or individualsReliability: lack of random error (consistency)Systematic ErrorSame effect on every measurementDoes not average outProduces confoundsAccuracy: lack of systematic errorDescriptive StatisticsMeasures of Central TendencyMeanMedianModeMeasures of VariabilityRangeStandard DeviationRelationship between two variables; CorrelationsScatterplotThe correlation coefficientStrength of relationship(0.0 to ±1.0)Direction of relationshipPositive vs negative correlationAssumes linear relationshipsNull vs. alternative hypothesesNull Hypothesis H0Default hypothesisStates that no effect exists0 correlation between 2 variables, or no difference between 2 groupsAlternative Hypothesis H1The “real” hypothesisStates that there is an effectAlways the logical opposite of H0Goal of most research is to reject H0 and allow adoption of H1StatisticsPopulation: the group of people/events a theory applies toSample: the people/events observed in a particular studyGoal is to learn about population, but we only have access to sampleRandom error in data means there will always be some effect in the sample, even when not present in populationInferential statisticsHelp determine whether effect is present in population (“real”) or simply due to chanceCompare observed effect to what is likely due to chance aloneRelies critically on amount of random error in dataSignificant vs. nonsignificant differencesp-value: probability that observed effect would occur by chance alone-level: cutoff used for p; usually 5%Statistically significant effectp < effect is bigger than would be likely by chancecan reject H0 and (tentatively) accept H1Non-significant effectp > Effect could have happened by chanceCannot reject H0; need further researchEvaluating researchIs there a theoretical basis for the work?Appropriate methods and levels of analysis?Adequate operational definitions?Appropriate evidence of causal relationship?Careful experimental design and control?To what populations do the results apply?Psych 301, 9/5/3Data AnalysisQuality of DataValidity–The extent to which the data collected measure the intended variableError–Effects due to variables other than those measured–NOT just noise or mistakes; includes e.g. variation among individuals–2 types: Random and SystematicRandom Error –Amount is different on each measurement–Tends to average out to zero over time or individuals–Reliability: lack of random error (consistency)Systematic Error –Same effect on every measurement–Does not average out–Produces confounds–Accuracy: lack of systematic errorDescriptive StatisticsMeasures of Central Tendency–Mean–Median–ModeMeasures of Variability–Range–Standard DeviationRelationship between two variables; CorrelationsScatterplot The correlation coefficient–Strength of relationship (0.0 to ±1.0)–Direction of relationshipPositive vs negative correlation–Assumes linear relationshipsNull vs. alternative hypothesesNull Hypothesis H0–Default hypothesis–States that no effect exists–0 correlation between 2 variables, or no difference between 2 groupsAlternative Hypothesis H1–The “real” hypothesis–States that there is an effect–Always the logical opposite of H0Goal of most research is to reject H0 and allow adoption of H1StatisticsPopulation: the group of people/events a theory applies toSample: the people/events observed in a particular studyGoal is to learn about population, but we only have access to sampleRandom error in data means there will always be some effect in the sample, even whennot present in populationInferential statistics–Help determine whether effect is present in population (“real”) or simply due to chance–Compare observed effect to what is likely due to chance alone–Relies critically on amount of random error in dataSignificant vs. nonsignificant differences–p-value: probability that observed effect would occur by chance alone–-level: cutoff used for p; usually 5%–Statistically significant effectp < effect is bigger than would be likely by chancecan reject H0 and (tentatively) accept H1–Non-significant effectp > Effect could have happened by chanceCannot reject H0; need further researchEvaluating researchIs there a theoretical basis for the work?Appropriate methods and levels of analysis?Adequate operational definitions?Appropriate evidence of causal relationship?Careful experimental design and control?To what populations do the results
View Full Document