You are on page 1of 18

Donation vs Discount

December 9, 2021
Kimberly Luu
Study Objectives
• Winterich, Carter, Barone, Janakiraman and Bezwada (2015)
studied the receptivity of individuals to promotions on
common products (bottled water, cereal) as a function of a
number of qualitative and quantitative independent
variables including gender, promotion type (donation to a
charity or discounted price), discount/donation size,
urban/rural residence, and personality trait
(interdependence).
• We will build an ANOVA model focusing only on their
bottled water purchase data: receptivity to purchase as a
function of discount type, discount level, and gender.
Activities
• Evaluate the nature of the data
• Construct the appropriate ANOVA probability
model
• Fit the model, validate the model, and test
hypotheses about its terms
• Make recommendations to the researchers based
on our findings
Data Summary
• The data set has 510 observations, each with the
following information:
• receptivity (qualitative ordinal: 1 = would not purchase,
7 = definitely purchase – we will treat it as quantitative)
• promotion type (qualitative nominal: donation or
discount)
• promotion level (qualitative ordinal: low, med, high)
• gender (qualitative nominal: 1 = female, 2 = male)
Research Questions
• Want to study which factors affect propensity to buy product.
• Does type of promotion matter? How much?
• Does level of promotion matter? How much? (The firm would like
to minimize this)
• Does the best promotion type depend on the gender of the
purchaser? How much?
• The effect from one factor may depend on the setting for another
– so interactions may be present.

• Appropriate probability model:


𝑌!"# = 𝜇 + 𝛼! + 𝛽" + 𝛾# + 𝛼𝛽!" + 𝛼𝛾!# + 𝛽𝛾"# + 𝜀!"#$ , 𝜀~𝑁(0, 𝜎 % ),

𝜇!"# ≡ 𝜇 + 𝛼! + 𝛽" + 𝛾# + 𝛼𝛽!" + 𝛼𝛾!# + 𝛽𝛾"# ,

1 𝛼! = 1 𝛽" = 1 𝛾# = 1 1 𝛼 𝛽!" = 1 1 𝛼 𝛾!# = 1 1 𝛽 𝛾"# = 0


Probability (ANOVA) Model
𝑌!"# = 𝜇 + 𝛼! + 𝛽" + 𝛾# + 𝛼𝛽!" + 𝛼𝛾!# + 𝛽𝛾"# + 𝜀!"#$ , 𝜀~𝑁(0, 𝜎 % ),

𝜇!"# ≡ 𝜇 + 𝛼! + 𝛽" + 𝛾# + 𝛼𝛽!" + 𝛼𝛾!# + 𝛽𝛾"# ,

1 𝛼! = 1 𝛽" = 1 𝛾# = 1 1 𝛼 𝛽!" = 1 1 𝛼 𝛾!# = 1 1 𝛽 𝛾"# = 0

• ai is the effect on receptivity when using the ith


promotion type, i = 1 or 2 (discount or donation)
• bj is the effect on receptivity when using the jth
promotion level, j = 2, 3 or 1 (med, high, or low)
• gk is the effect on receptivity for the kth gender, k =
1 or 2 (female or male)
Box Plot
• Gender makes greatest difference
• Donation seems to work better than discount, maybe
more so for females
Minitab Model Specification
Minitab Model Specification
Hypotheses and Significance
Null and Alternate :
• H0: all alpha, beta, gamma, alphabeta,
alphagamma, betagamma coeffcients = 0
• H1: at least one coefficient not = 0

Level of Significance:
• Level of significance = 0.05
• Typical for business decisions
ANOVA Model Results
• Only gender shows up as important, with a p-value
< 0.5
ANOVA Model Results
• Very low 𝑅 2 , percent variation explained only 3%
ANOVA Model Results
• Gender coefficient is the largest, but not even a full
point on the Likert scale.
Residuals Plots
• Significant
problems
• Nonlinear
normal plot
• Non-bell shaped
histogram
• Strange ‘vs fits’
plot
• We cannot trust
the p-values
ANOVA Model Residual Plots and
ANOVA with Ordinal Y
• We would not expect these plot to be good:
• From http://www.pmean.com/09/LikertAnova.html
“I never quite feel I can offer my students a thoughtful explanation about
the use of Likert data with ANOVA. It is recommended that ANOVA be
used with interval or ratio data, but, in practice, ANOVA is sometimes used
when the data is ordinal (as you'd find when using Likert scales). This
confuses some students. Are there any good references out there I can
share with my students that might explain the pros and cons of using
ordinal data with ANOVA?

Your discomfort is shared by many of us. I have vacillated on this issue


many times. I would just let students know that use of ANOVA for Likert
scale items is controversial. There is no consensus in the research
community on how to handle this type of data. When there is no
consensus, choose whatever you like, but be prepared to re-analyze the
data when the peer-reviewer asks you to change to the competing
procedure.
ANOVA with Ordinal Y
• Means that p-values cannot be trusted
• But coefficient estimates are best (least squares)
available
Study Summary and
Recommendations
• Although we measured whether three different variables
contributed to the propensity to buy a product, gender had
the highest propensity.
• Used residuals plots to support our finding of p-value. We
found that it was not trustworthy.
• Our next best estimate would be to look at the coefficient,
however, the greatest coefficient (gender) was still relatively
low.
• The promotion type and level does not seem to matter since
their p-values are relatively larger compared to gender.
• Their coefficients are also lower compared to gender’s coefficient.
• The interaction between promotion type and gender also
has a high p-value and low coefficient, so there may not be
a relationship.

You might also like