Deck 15: Testing for Association: Phi

Question 1

One of the reasons that Phi is a popular statistic is that it has both symmetric and asymmetric versions.

Answer 1

False

Question 2

If a two-variable table creates problems for Chi-square, there will also be problems with Phi.

Answer 2

True

Question 3

Phi is based on an inferential statistic; that statistics is

A) Analysis of Variance
B) t-test
C) Chi-square
D) Fisher's Exact Test

Answer 3

C

Question 4

The Uncertainty Coefficient is possible statistic to use with nominal level variables when there are serious table problems.

Question 5

Phi is a correlation statistic for which level of measurement below?

A) nominal
B) ordinal
C) interval
D) nominal and ordinal

Question 6

[Instructor note: Use this question if you are combining questions on measures of association.] An appropriate measure of association for determining the strength of the relationship between religious affiliation (Protestant, Other) and gender (Male, female) is:

A) Somers' d
B) Phi
C) chi-square
D) Pearson's r

Question 7

A Phi value of .40 would be interpreted as having explained

A) 4% of the variance in the dependent variable
B) 40% of the variance in the dependent variable
C) 1.6% of the variance in the dependent variable
D) 16% of the variance in the dependent variable

Question 8

Correlation coefficients for Phi range between zero and 1.0.

Question 9

In order to interpret Phi, you must

A) divide the correlation coefficient by N
B) square the correlation coefficient
C) multiply the correlation coefficient by 2
D) multiply the correlation coefficient by 100

Question 10

Smaller tables, on the whole, are easier to interpret at the nominal level than larger tables.

Question 11

Under which of these circumstances would you prefer NOT to use Phi?

A) the number of cases is large
B) the relationship is small
C) the coefficient produces a value less than .50
D) the table marginals are very different

Question 12

The nominal-level concept of association is that

A) as the number of cases in the table cells of one of the values increases, the corresponding number should decrease in cells of the other values
B) as the number of cells in the row of the independent variable increases, the corresponding number of cells should decrease in cells of the other rows
C) the rank of scores in the cells of the rows should match the rank of scores in the cells of the columns
D) all of the above

Question 13

There are alternatives to the use of Phi; two of them are:

A) Gamma and Lambda
B) Uncertainty Coefficient and Spearman's Rho
C) Chronbach's Alpha and McNemar's Q
D) none of the above

Question 14

A Phi coefficient of .6 means that the percent of explained variance is 36%.

Question 15

A perfect relationship, as defined by Phi, is

A) all cases in one row or column
B) all cases in two rows or columns
C) zero cases on the table diagonal
D) all cases on the table diagonal

Question 16

Assumptions made by the Phi statistic include

A) normal distributions
B) equal variances
C) variables are natural dichotomies
D) all of the above

Question 17

You can't tell the direction of a relationship by looking only at its Phi coefficient.

Question 18

Phi is designed to be used with square tables.

Question 19

If you explained 25% of the variance in the dependent variable using the Phi statistic, what would the Phi coefficient be?

A) .25
B) .40
C) .50
D) none of the above

Question 20

Which of the variables below would be interpreted as a natural dichotomy?

A) age measured by "young" and "old"
B) gender measured by "male" and "female"
C) income measured by "low" and "high"
D) type of crime measured by "nonserious" and "serious"
Testing for Association: Phi

Question 21

Phi is sensitive to unequal marginal distributions.

Question 22

Any variable that can be split into two equal parts is, by definition, a natural dichotomy.