Abstract
This chapter introduces ANOVA, (Analysis of Variance) to analyze data from conjoint analysis experiments. Conjoint analysis is used in experiments to quantify customer preferences for better design of new products and services.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Author information
Authors and Affiliations
9.1 Electronic Supplementary Material(s)
Appendices
JMP 9.1: Run ANOVA and Pairwise Comparisons of Factor Levels
Fit Model in JMP is a General Linear Model. If potential drivers are categorical, output will be ANOVA with paired comparisons.
In beer taste.xlsx, copy and paste columns A through D into a new JMP data file.
Analyze, Fit Model.
Drag rating to Pick Role Variables Y.
Drag the three factors, brewery, alcohol and flavor to Construct Model Effects.
Set Emphasis to Minimal Report.
Run.
Summary of fit contains multiple regression results with indicators set up by JMP. (JMP will do this if drivers are recognized as ordinal or nominal and not continuous.)
Effect Tests presents the factor F tests and pvalues.
All three factors make a significant difference in ratings.
Click on the triangle next to Effect Details to see the factor level means.
For each factor, click on the red triangle next to the factor, such as brewery, and choose LSMeans Student’s t to see pairwise comparisons with t tests and pvalues.
Significant differences are shown in red. All brewery pairs are significantly different. Below the table, the part worths are sorted in descending order. Equivalent part worths share the same letter. CA micro is preferred to the other brewery levels, and VA micro is also preferred to US name brand.
6% and 4% differ from 8%, but not from each other. Either 6% or 4% are preferred to 8%.
All pairwise comparisons of flavor options differ significantly. Lime is preferred to the two alternate flavors, and neither is preferred to coffee flavor.
For each factor, click on the red triangle next to the factor, such as brewery, and choose LSMeans plot to see a plot of the factor level means.
To export a copy of your output, under File, choose Export:
Then, choose Microsoft Word or PowerPoint, Next, and enter the file location:
In the Word or PowerPoint file, resize the plots to display side by side and add stand alone titles:
Copy and paste the Analysis of Variance and Effect Tests from the word document into the excel document.
Create a pie graph to illustrate factor contributions to explanation of preference variation (RSquare) from the Sums of Squares, shown in Fig. 9.3:
Right click inside a slice to select Add Data Labels. Then right click on a data label to Format Data Labels, choosing Category Name, Percentage, and unchecking Value.
Delete the legend and make the title stand alone, shown in Fig. 9.4. Click a slice twice to change the fill color in Format, if you choose.
Case 9.1: Background Music to Enhance Ad Message Recall
A brand manager suspects that the background music featured in a brand’s advertising may affect consumers’ recall of information presented in ads. Some music backgrounds are more distracting than others, and may compete with audience attention to the advertising message. Several background options are being considered, and those options differ along three categories, or factors.
-
Three vocals options are: backgrounds which feature vocals, backgrounds with brand related vocals substituted for original vocals, and backgrounds with vocals removed.
-
Three orchestration options are: saxophone, saxophone and percussion, and saxophone and piano.
-
Three tempo options are: slow tempo, moderate tempo, and fast tempo
It has been established that multiple timbres (from multiple instruments) distract more. It is also known that changes in the music background distract, and the rate of distraction is lower for slower tempos; yet, faster tempos allow streaming – music changes heard before become less surprising and distracting at faster tempos.
To determine whether vocals, orchestration and tempo of backgrounds affect brand message recall, the ad agency creative team designed nine backgrounds for a brand ad using conjoint analysis. Since the ad message, visuals, and length of ad could also influence message recall, the agency creatives were careful to make those ad features identical across the nine versions. By using ads that were identical, except for their musical backgrounds, any difference in resulting brand message recall could be attributed to the difference in backgrounds. The conjoint analysis design is shown below:
Orchestration | None | Original | Brand specific |
Sax | Slow | Moderate | Fast |
Sax & percussion | Moderate | Fast | Slow |
Sax & piano | Fast | Slow | Moderate |
Eighteen consumers were randomly selected and then randomly assigned to one of the nine background treatments, or combination of vocals, orchestration and tempo. Each viewed the brand advertisement with one of the nine backgrounds, and then message elements in the ad, which could be six, if all elements were recalled, or as low as zero, if no elements were recalled.
The data are in music backgrounds.xlsx.
Test hypotheses regarding factor effects on message recall
-
1.
State the null hypotheses for the three factors. Indicate which you can reject, specifying the appropriate statistic, degrees of freedom and pvalues for all three.
-
2.
Assess pairwise differences between factor levels
Which vocal(s) lead to higher recall?
___ none ___ original ___ brand specific
Which orchestration(s) lead to higher recall?
___ sax ___ sax and percussion ___ sax and piano
Which tempo(s) lead to higher recall?
___ slow ___ moderate ___ fast
-
3.
Plot the part worth recall scores by factor level for each of the three factors with a line graph.
-
4.
Plot the sums of variance explained by each factor in a pie chart.
Summarize what you learned from your analysis:
Rights and permissions
Copyright information
© 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Fraser, C. (2024). Conjoint Analysis and Experimental Data. In: Business Statistics for Competitive Advantage with Excel and JMP . Springer, Cham. https://doi.org/10.1007/978-3-031-42555-4_9
Download citation
DOI: https://doi.org/10.1007/978-3-031-42555-4_9
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-42554-7
Online ISBN: 978-3-031-42555-4
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)