Correlational Methods and Statistics
Correlation Nonexperimental method that describes a relationship between two variables. Allow us to make predictions If smoking is correlated with lung cancer, then we can predict, with some accuracy, that a person who smokes can develop lung cancer. Used when: Unethical to conduct experimental study (i.e., smoking condition) Researchers want to assess the relationship among many variables at once. Ex: variables that correlate with personality traits
Characteristics of Correlations Magnitude strength of the relationship measured by a correlation coefficient weak moderate strong no relationship
How strong is an association?
Scatterplots Graphical representation of the relationship between 2 variables. r = -.60
Curvilinear relationship No Correlation Negative Correlation Positive Correlation
Interpretation Difficulties Only experiments allow us to infer causality and directionality factor A caused factor B to change. Correlational studies no inferences of causality or directionality “Correlation does not imply causation” be a critical consumer of information
Third- Variable Problem Despite strong correlation, the results could be due to something else… Third-variable problem: the correlation between 2 variables is dependent on another variable. Ex: teenage delinquency increases with sales of ice-cream
Restrictive Range Occurs when a variable has limited variability due to restrictions in range exposure to noise (months) exposure to noise (years) Hearing ability
Pearson’s r Duration of Cold symptoms Hours of sleepzColdzHours zCold*zHour s Mean = 4.73Mean = 5.53N = 15SUM = SD = 1.79SD = 1.46r = Population r = ∑ (zA)(zB) __________ N Sample r = ∑ (zA)(zB) __________ N - 1
Alternative correlation coefficients range of coefficients: -1, 0, + 1 Spearman’s rank order correlation coefficient both variables are ordinal (ranked) Point-Biserial one variable is interval or ratio other variable is nominal (and has only 2 levels; ex: gender) Phi both variables are nominal and have only 2 levels in each.
Regression Analysis Procedure that allows us to predict an individual’s performance on variable A from knowing variable B. Determines the best-fitting line for a data set. Y’ = bX + a Y’ is the predicted value b is the slope of a line X is the subject’s score a is the Y-intercept
Multiple Regression Combines several predictor variables into one regression equation. Allows for access of the effects of multiple predictor variables on a dependent measure. Represented by “R” Ex: smoking influences the likelihood of developing cancer, but other factors like genetic predisposition, life style and nutrition can help us predict cancer development.