What type of test

1
What type of test?

• A college recruiter asks potential college
  students which of four universities in WA state
  they would prefer to attend. The data are shown
  below.
• UW (n 10), WSU (n 20), WWU (n 5), EWU (n
  5)
• Answer? Chi-square This is a nominal scale, and
  these data are frequencies

2
What type of test?

• A bakery with locations at different shopping
  malls across the U.S. is trying to increase
  sales. At one store, they do nothing. At another
  store, they pump the smell of cookies into the
  mall. At another store, they pump the smell of
  cookies into the mall and provide free samples of
  the cookies. On average, they find that people
  buy 1 cookie at the first type of store, they buy
  2 cookies at the second type of store, and they
  buy 5 cookies at the third type of store.
• Answer? ANOVA. You have three group of subjects
  (customers at the three different stores), and
  they give you responses on a continuous scale (
  of cookies purchased).

3
What type of test?

• An electronics store wants to know whether people
  prefer the iPod classic or the iPod nano. They
  ask 25 people to rate the iPod (1 I hate it, 10
  I love it). They ask another 20 people to rate
  the iPod nano on the same scale.
• Answer? Independent t-test. You have two separate
  (independent) groups, and you measure each
  subjects rating of the iPod, and then
  (presumably) compare the means.

4
What type of test?

• A researcher is interested in determining whether
  people have a favorable attitude toward a certain
  brand of toothpaste. She asks 20 people to rate
  how much they like the toothpaste on a scale from
  1 (hate it), 4 (neutral), 7 (love it). The
  average response is 4.7, with a standard
  deviation of 1. Do people like this toothpaste
  (score significantly above the scale midpoint)?
• Answer? Single sample t-test. Here you have a
  sample mean (mean 4.7), a sample standard
  deviation (s 1), a sample size (n 20) and a
  point of comparison (scale midpoint of 4).

5
What type of test?

• A researcher is interested in whether the outside
  temperature has any impact on peoples belief in
  global warming. Presumably, the warmer it is
  outside, the more people will agree that global
  warming is taking place. She asks people to
  complete a global warming scale (1 its a hoax,
  100 its a real problem) and later goes back
  and records the outside temperature for the day
  the subject completed the survey.
• Answer? Correlation. You are looking at a
  relationship between two continuous variables
  (temperature and beliefs in global warming). You
  could also use regression analysis if you wanted
  to predict a persons belief in global warming
  based on the outside temperature.

6
What type of test?

• A research firm is hired to determine the impact
  of music on clothing sales. In one store, they
  play classical music. In another store, they play
  classic rock. The firm takes sales data on the
  first 100 customers to enter each store and then
  compares the average amount spent in the two
  stores.
• Answer? Independent t-test. You have two separate
  (independent) groups, and you measure how much
  each of the 100 subjects in each store spends on
  clothes.

7
What type of test?

• A candy store is interested in whether people
  prefer dark chocolate or milk chocolate. They ask
  30 customers to try some dark chocolate and then
  rate it on a 7-point scale (1 hate it to 7
  love it). Then they ask the same customers to try
  some milk chocolate and then rate it on the same
  7-point scale. Later, they compare the means for
  the two conditions.
• Answer? Paired samples t-test. You have one group
  of subjects, and they are tested under two
  different conditions (dark vs. milk chocolate).

8
What type of test?

• A professor is interested in the relationship
  between consumer confidence and spending during
  the Christmas season. He surveys 100 people and
  asks them to fill out a scale assessing their
  confidence in the economy and how much they spent
  during Christmas.
• Answer? Correlation. You have one group of
  subjects, and they provide two measures
  (confidence and spending), and the professor is
  looking for a relationship between them. You
  could also use regression analysis if you wanted
  to predict a persons spending based on their
  confidence.

9
What type of test?

• A new home furnishings store is trying to decide
  what color to paint the walls. They ask 60 people
  to evaluate one of three mock ups of the store.
  Some see the store in green, some see the store
  in blue, and some see the store in orange
  (pumpkin to be precise). Subjects rate the store
  on a scale from 0 (awful) to 100 (wonderful). The
  means are (green 58, blue 80, orange 50).
• Answer? ANOVA. You have three group of subjects,
  and they give you responses on a 100-point scale.

10
What type of test?

• A music store is trying to figure out whether it
  should begin stocking DVDs of concerts. They call
  up 200 people selected at random and ask them if
  they would be interested in purchasing this type
  of DVD if stocked. 60 say yes.
• Answer? Chi-square This is a nominal scale (yes,
  no), and these data are frequencies. Technically,
  all you see is 60, so that makes this question
  tricky. But its really not tricky at all. All
  you need to do is convert 60 to the actual
  frequencies Yes (60 x 200 120), No (40 x 200
  80). So, in the end, you do have frequencies.

11
What type of test?

• Coke is interested in whether people prefer the
  old classic coke or the new coke. They ask 20
  people to sample both types of coke and rate them
  on a scale from 1 (hate it) to 7 (love it). They
  then compare the means for new and old coke.
• Answer? Paired samples t-test. You have one group
  of people and they are assessed under two
  different conditions.

12
What type of test?

• Last year, the American Cancer Society reported
  an average contribution of 20. This year they
  ran a new campaign and recomputed the average
  contribution based on a survey of 1000 local
  donors. This year, the average was 25, with a
  standard deviation of 5.
• Answer? Single sample t-test. Here you have a
  sample mean (mean 25), a sample standard
  deviation (s 5), a sample size (1000), and a
  point of comparison (last years contribution of
  20).