Page 187 - Statistics II for Dummies

Chapter 9: Testing Lots of Means? Come On Over to ANOVA!

Upfront rejection is the best policy 171

for most refusal letters

Many medical and psychological studies the rejection letters) on some quantitative vari-
use designed experiments to compare the able (response to the letter). You can argue that
responses of several different treatments, look- response to the letter isn’t a continuous vari-
ing for differences. A designed experiment is a able, however it has enough possible values
study in which subjects are randomly assigned that ANOVA isn’t unreasonable. The data were
to treatments (experimental conditions) and also shown to have a bell shape.
their responses are recorded. The results are The null hypothesis would be Ho: Mean responses
used to compare treatments to see which to the three types of rejection letters are equal
one(s) work best, which ones work equally well, versus Ha: At least two forms of the rejection
and so on.
letter resulted in different mean responses.
Ohio State University researchers conducted
one such experiment using ANOVA to deter- In the end, the researchers did find some sig-
mine the most effective way to write a rejection nificant results; the different ways the rejec-
letter. (Is there really a best way to say “no” to tion letter was written affected the participants
someone? Turns out the answer is “yes.”) The differently (so the F-test was rejected). Using
experiment tested three traditional principles of multiple comparison procedures (see Chapter
writing refusal letters: 10), you could go in and determine which forms
of the rejection letters gave different responses
✓ Using a buffer, which is a neutral or posi- and how the responses differed.
tive sentence that delays the negative In case you have to write a rejection letter at
information
some point, the researchers recommend the
✓ Placing the reason before the refusal following guidelines:
✓ Ending the letter on a positive note as a way ✓ Don’t use buffers to begin negative
of reselling the business messages.
Subjects were randomly assigned to treat- ✓ Give a reason for the refusal when it makes
ments, and their responses to the rejection the sender’s boss look good.
letters were compared (likely on some sort of
scale such as 1 = very negative to 7 = very posi- ✓ Present the negative positively but clearly;
tive with 4 being a neutral response). offer an alternative or compromise if
possible.
You can analyze this scenario by using ANOVA ✓ A positive ending isn’t necessary.
because it compares three treatments (forms of

7/23/09 9:31:31 PM
15_466469-ch09.indd 171
15_466469-ch09.indd 171 7/23/09 9:31:31 PM