MORE COMPARISONS OF MEANS • You have more than two groups • and a mean (average) for each – e.g., young = 4.0, – middle aged = 5.0, – older = 4.5 • How do you determine the strength of the covariation? 7/17/2016 Marketing Research 1 ANOVA – Decomposes “variance” into: • treatment effects • other factors • unexplained factors – Compares data to group means • Subtracts each data point from group mean • Squares it • Keeps a running total of “Sum of Squares” 7/17/2016 Marketing Research 2 ANOVA – The Sums of Squares are then: • Divided by the number of groups • (To get an estimate “per group”) – “Mean Squares” • MSSr = SSr / df – (variance per group) • MSSr / MSSu = F – Total variance “explainable” – F compared to F crit [dfn, dfd] – if F > F crit, difference in population 7/17/2016 Marketing Research 3 ANOVA (continued) – One way ANOVA investigates: – Main effects • factor has an across-the-board effect • e.g., age • or involvement 7/17/2016 Marketing Research 4 Example • Study of movie profits – Dependent variable: • Gross revenue in dollars [continuous] – Independent variables: • Sex [categorical] • Violence – Examine predictors of profitability: • Sex, violence, interaction (sex * violence) 7/17/2016 Marketing Research 5 Main effect: Sex 5 4 No sex Sex 3 2 Low 7/17/2016 VIOLENCE LEVEL Marketing Research High 6 Main effect: Violence 5 4 No sex Sex 3 2 Low 7/17/2016 VIOLENCE LEVEL Marketing Research High 7 ANOVA – A TWO-WAY ANOVA investigates: – INTERACTIONS • effect of one factor depends on another factor • e.g., larger advertising effects for those with no experience • importance of price depends on income level and involvement with the product 7/17/2016 Marketing Research 8 Interaction: Sex by Violence 5 4 No sex Sex 3 2 Low 7/17/2016 VIOLENCE LEVEL Marketing Research High 9 Example • Study of movie profits – Dependent variable: • Gross revenue in dollars [continuous] – Independent variables: • Sex [categorical] • Violence – Examine predictors of profitability: • Sex, violence, interaction (sex * violence) 7/17/2016 Marketing Research 10 SPSS Output Tests of Between-Subjects Effects Dependent Variable: Total Gross Type III Sum Source of Squares Corrected Model 43744.364 a Intercept 952785.362 SEX 35467.649 VIOLENCE 10228.369 SEX * VIOLENCE 21.589 Error 995088.361 Total 1991539.265 Corrected Total 1038832.725 a. R Squared 7/17/2016 df 3 1 1 1 1 381 385 384 Mean Square 14581.455 952785.362 35467.649 10228.369 21.589 2611.780 = .042 (Adjusted R Squared .035) Marketing = Research F 5.583 364.803 13.580 3.916 .008 Sig. .001 .000 .000 .049 .928 11 MULTIVARIATE STATISTICS Previous statistics: • Crosstabs (chi-square) • t-test (means) • Analysis Of Variance (ANOVA) • Pearson’s correlation coefficient • Regression • Multiple regression 7/17/2016 Marketing Research 12 Next couple of statistics: • used less frequently • (bottom of your “tool box”) • more “exploratory” 1. Discriminant analysis 2. Factor analysis 3. Cluster analysis 7/17/2016 Marketing Research 13 1. Discriminant analysis • Tests for covariation between: • Categorical dependent variable – “group” such as purchaser, non purchaser; – regular, occasional, and infrequent purchasers; – people who like “chick” flicks, those that like “guy flicks,” those that like both • Continuous and categorical IVs – age, income, gender, etc. 7/17/2016 Marketing Research 14 1. Discriminant analysis (continued) • which variables “group” people? • coefficients reveal importance of factors – larger coefficient, more important – smaller coefficient, less important • p-value associated with specific variables • overall fit assessed by “% correctly classified” 7/17/2016 Marketing Research 15 2. Factor analysis • examines covariation between: • several different variables • that are “reduced” to one or more underlying “factors” or “constructs” – e.g., overall “intelligence,” “need to consume,” etc. • often used to develop scales of related questions -- “data reduction” • no information on what causes what 7/17/2016 Marketing Research 16 2. Factor analysis (continued) • attempts to identify the most shared variation – the first factor is the largest amount of “variance” – the second factor is the second largest variance, etc. • Number of “factors” – “Eigenvalues” > 1.0 (“chance”) – “bend in the “Scree plot” 7/17/2016 Marketing Research 17 3. Cluster analysis • Covariation among a number of variables • Identifies “segments” of the sample • Used frequently in marketing • Helpful for targeting products and messages 7/17/2016 Marketing Research 18 4. Multi-Dimensional Scaling • abbreviated “MDS” • Subject rate “distances” or differences between objects • Data are subjected to analysis • Analysis reveal underlying “dimensions” • Used to identify how people differentiate products 7/17/2016 Marketing Research 19 The End 7/17/2016 Marketing Research 20