Deck 19: Categorical Outcomes: Chi-Square and Loglinear Analysis

Question 1

What data transformations are necessary before loglinear analysis?

A) Data reduction
B) Logarithmic
C) Linear regression
D) Summary statistics

Answer 1

Logarithmic

Question 2

A study undertaken by Reed (2004) looked at an evaluation of the effect of an intervention across medical care providers in an institution and at physician group practice level. The patient group were diabetics. Within the physician care group, there were a non-intervention group of patients and an intervention group. The following result was derived using the Pearson chi-square: χ² = 99.58, p = .0001 (source: http:/ / www.annfammed.org/ cgi/ reprint/ 2/ 3/ 201.pdf). What does this suggest about the result?

A) No difference between the groups
B) A significant difference between the groups and intervention effective
C) A significant difference between the groups and intervention not effective
D) None of the above

Answer 2

A significant difference between the groups and intervention effective

Question 3

A calculated value of chi-square

A) Compares the frequencies of categories of items in a sample to the frequencies that are expected in the population.
B) Compares the frequencies of categories in the population to those expected in the sample.
C) Compares the mean number of categories in a sample to those in the population.
D) Compares the mean number of frequencies in a population to those in the sample.

Answer 3

Compares the frequencies of categories of items in a sample to the frequencies that are expected in the population.

Question 4

Which of the following are not examples of assumptions for the chi-square statistic?

A) Each entity contributes to only one cell of the contingency table.
B) Expected frequency should be less than 5.
C) It does not rely on having continuous normally-distributed data.
D) None of the above.

Question 5

Which of the following is not an assumption to be met for loglinear analysis?

A) Cells of the contingency table must be independent.
B) All cells must have expected frequencies greater than 5.
C) All cells must have expected frequencies greater than 1.
D) All of the above.

Question 6

In which sub-dialog box can the chi-square test be found?

A) Frequencies: Percentages
B) Crosstabs: Statistics
C) Bivariate: Pearson
D) sex: Female

Question 7

In the following table, what number is missing? (marked with ?)

A) 6
B) 2.6
C) 1.4
D) 9.4

Question 8

Chi-square is a test of

A) Difference
B) Relationship
C) Association
D) Factors

Question 9

Which is an assumption of chi-square?

A) That the data are normally distributed.
B) That each subject contributes data to only one cell.
C) That the data are continuous.
D) That there is little variability among the data.

Question 10

Loglinear analysis is

A) A multivariate alternative to chi-square.
B) A multivariate equivalent to linear regression.
C) An analytical form of logistic regression.
D) A way of working with logarithms when there are a lot of numbers.

Question 11

Under what circumstances would Yates's correction be used?

A) Large quantities of data
B) In 2 × 2 tables of data
C) In 3 × 3 tables of data
D) In 4 × 4 tables of data

Question 12

Loglinear analysis is not

A) A goodness-of-fit test that allows you to test all the effects.
B) A test that employs an algorithm to generate expected cell frequencies for each model .
C) A test that will find the model that best represents the data.
D) A test that employs a differential analysis.

Question 13

Which of the following is an advantage of loglinear analysis?

A) It allows analysis of different types of data.
B) It provides a systematic approach to the analysis of complex multidimensional tables.
C) Empty cells do not lower the power of the test.
D) It means you do not have to compare models.

Question 14

When testing several categorical variables, which is the appropriate test of relationship?

A) Chi-square
B) Elaborated chi-square
C) Part correlation
D) Loglinear analysis

Question 15

Kuhn et al. (1997) identify the following Pearson chi-square test result when looking at the shape of acromion and rotator cuff tears: Х² = 17.636, P = .0005 (source: http:/ / www.sjsu.edu/ faculty/ gerstman/ EpiInfo/ cat-sev.htm). What would you deduce from this result?

A) The result is significant.
B) The result is not significant.
C) The result needs to be modified using Fisher's exact test.
D) None of the above.

Question 16

In a study by Sheiham et al. (2002) of 'the relationship between oral health status and body mass index among older people' in Great Britain, Fisher's exact test was used (source: http:/ / www.nature.com/ bdj/ journal/ v192/ n12/ full/ 4801461a.html). Why would this test have been used in this instance?

A) An adjusted odds ratio would not provide a valid effect size.
B) An adjusted odds ratio could not be used as the sample was too large.
C) An adjusted odds ratio could not be used as the sample was too small.
D) none of the above.

Question 17

For the same data, a chi-square test produced the SPSS output below. What can we conclude from this output?

A) Women watched significantly more programmes than men.
B) Significantly more soap operas were watched.
C) The profile of programmes watched was significantly different between men and women.
D) Men and women watch similar types of programmes.

Question 18

A researcher asked 933 people which type of programme they prefer to watch on television. Results are below. What is the expected frequency for men who liked to watch sport?

A) 132
B) 94.19
C) 64.09
D) 27.45

Question 19

When would you consider converting χ² to odds ratio?

A) When calculating the effect size before performing a chi-square test
B) When calculating the likelihood before performing a chi-square test
C) When calculating the effect size after performing a chi-square test
D) When calculating the likelihood after performing a chi-square test

Question 20

Based on the data above, what are the odds of being a man if you watch sport?

A) 0.47
B) 0.14
C) 0.41
D) 2.13

Question 21

Which of the following is an advantage of loglinear analysis?

A) It allows analysis of different types of data.
B) It provides a systematic approach to the analysis of complex multidimensional tables.
C) Empty cells do not lower the power of the test.
D) It means you do not have to compare models.

Question 22

Which of the following is not an assumption of loglinear analysis?

A) An entity should fall into only one cell of the contingency table.
B) Expected frequencies should be large enough for a reliable analysis.
C) If there are two variables 20% of cells must have frequencies less than 1.
D) Collapsed data need to have a non-significant highest-order interaction.

Question 23

A study undertaken by Reed (2004) looked at an evaluation of the effect of an intervention across medical care providers in an institution and at physician group practice level. The patient group were diabetics. Within the physician care group, there were a non-intervention group of patients and an intervention group. The following result was derived using the Pearson chi-square: χ² = 99.58, p = .0001 (source: http:/ / www.annfammed.org/ cgi/ reprint/ 2/ 3/ 201.pdf). What does this suggest about the result?

A) No difference between the groups
B) A significant difference between the groups and intervention effective
C) A significant difference between the groups and intervention not effective
D) None of the above

Question 24

When would you consider converting χ² to odds ratio?

A) When calculating the effect size before performing a chi-square test
B) When calculating the likelihood before performing a chi-square test
C) When calculating the effect size after performing a chi-square test
D) When calculating the likelihood after performing a chi-square test

Question 25

Based on the data above, what are the odds of being a man if you watch sport?

A) 0.47
B) 0.14
C) 0.41
D) 2.13

Question 26

When testing several categorical variables, which is the appropriate test of relationship?

A) Chi-square
B) Elaborated chi-square
C) Part correlation
D) Loglinear analysis

Question 27

In which sub-dialog box can the chi-square test be found?

A) Frequencies: Percentages
B) Crosstabs: Statistics
C) Bivariate: Pearson
D) sex: Female

Question 28

A researcher asked 933 people which type of programme they prefer to watch on television. Results are below. What is the expected frequency for men who liked to watch sport?

A) 132
B) 94.19
C) 64.09
D) 27.45

Question 29

A calculated value of chi-square

A) Compares the frequencies of categories of items in a sample to the frequencies that are expected in the population.
B) Compares the frequencies of categories in the population to those expected in the sample.
C) Compares the mean number of categories in a sample to those in the population.
D) Compares the mean number of frequencies in a population to those in the sample.

Question 30

Which of the following is not an assumption to be met for loglinear analysis?

A) Cells of the contingency table must be independent.
B) All cells must have expected frequencies greater than 5.
C) All cells must have expected frequencies greater than 1.
D) All of the above.

Question 31

Which is an assumption of chi-square?

A) That the data are normally distributed.
B) That each subject contributes data to only one cell.
C) That the data are continuous.
D) That there is little variability among the data.

Question 32

What data transformations are necessary before loglinear analysis?

A) Data reduction
B) Logarithmic
C) Linear regression
D) Summary statistics

Question 33

When would you consider using a loglinear analysis as the method of choice?

A) When data need to be collapsed
B) When sample sizes are large
C) When there are several categorical variables
D) None of the above

Question 34

Chi-square is a test of

A) Difference
B) Relationship
C) Association
D) Factors

Question 35

For the same data, a chi-square test produced the SPSS output below. What can we conclude from this output?

A) Women watched significantly more programmes than men.
B) Significantly more soap operas were watched.
C) The profile of programmes watched was significantly different between men and women.
D) Men and women watch similar types of programmes.

Question 36

Loglinear analysis is

A) A multivariate alternative to chi-square.
B) A multivariate equivalent to linear regression.
C) An analytical form of logistic regression.
D) A way of working with logarithms when there are a lot of numbers.

Question 37

In the following table, what number is missing? (marked with ?) SEX * GROUPS Crosstabulation

A) 6
B) 2.6
C) 1.4
D) 9.4

Question 38

A study entitled 'Structural equation and log-linear modelling: a comparison of methods in the analysis on caregivers' health' by Zhu et al. (2006) showed a large effect size in the interaction between psychological health and physical health (source: http:/ / www.biomedcentral.com/ 1471-2288/ 6/ 49). What can be deduced by this finding?

A) Both factors are significantly related in the study.
B) Both factors are significantly unrelated in the study.
C) Both factors have an effect on the study results.
D) None of the above.

Question 39

In a study of 'communicating the significance of radiologic test results' by Black and Armstrong (1986), the use of likelihood ratios is very important to the work of the radiologist (source: http:/ / www.ajronline.org/ cgi/ reprint/ 147/ 6/ 1313). Why?

A) Because it provides a reliable predictor of disease probability.
B) Because it assists with computation of the result.
C) Because it has a low predictive power.
D) None of the above.

Question 40

Loglinear analysis is not

A) A goodness-of-fit test that allows you to test all the effects.
B) A test that employs an algorithm to generate expected cell frequencies for each model .
C) A test that will find the model that best represents the data.
D) A test that employs a differential analysis.

Question 41

In a study of 'communicating the significance of radiologic test results' by Black and Armstrong (1986), the use of likelihood ratios is very important to the work of the radiologist (source: http:/ / www.ajronline.org/ cgi/ reprint/ 147/ 6/ 1313). Why?

A) Because it provides a reliable predictor of disease probability.
B) Because it assists with computation of the result.
C) Because it has a low predictive power.
D) None of the above.

Question 42

Which of the following is not an assumption of loglinear analysis?

A) An entity should fall into only one cell of the contingency table.
B) Expected frequencies should be large enough for a reliable analysis.
C) If there are two variables 20% of cells must have frequencies less than 1.
D) Collapsed data need to have a non-significant highest-order interaction.

Question 43

Kuhn et al. (1997) identify the following Pearson chi-square test result when looking at the shape of acromion and rotator cuff tears: Х² = 17.636, P = .0005 (source: http:/ / www.sjsu.edu/ faculty/ gerstman/ EpiInfo/ cat-sev.htm). What would you deduce from this result?

A) The result is significant.
B) The result is not significant.
C) The result needs to be modified using Fisher's exact test.
D) None of the above.

Question 44

Which of the following are not examples of assumptions for the chi-square statistic?

A) Each entity contributes to only one cell of the contingency table.
B) Expected frequency should be less than 5.
C) It does not rely on having continuous normally-distributed data.
D) None of the above.

Question 45

A study entitled 'Structural equation and log-linear modelling: a comparison of methods in the analysis on caregivers' health' by Zhu et al. (2006) showed a large effect size in the interaction between psychological health and physical health (source: http:/ / www.biomedcentral.com/ 1471-2288/ 6/ 49). What can be deduced by this finding?

A) Both factors are significantly related in the study.
B) Both factors are significantly unrelated in the study.
C) Both factors have an effect on the study results.
D) None of the above.

Question 46

When would you consider using a loglinear analysis as the method of choice?

A) When data need to be collapsed
B) When sample sizes are large
C) When there are several categorical variables
D) None of the above

Question 47

Under what circumstances would Yates's correction be used?

A) Large quantities of data
B) In 2 × 2 tables of data
C) In 3 × 3 tables of data
D) In 4 × 4 tables of data

Question 48

In a study by Sheiham et al. (2002) of 'the relationship between oral health status and body mass index among older people' in Great Britain, Fisher's exact test was used (source: http:/ / www.nature.com/ bdj/ journal/ v192/ n12/ full/ 4801461a.html). Why would this test have been used in this instance?

A) An adjusted odds ratio would not provide a valid effect size.
B) An adjusted odds ratio could not be used as the sample was too large.
C) An adjusted odds ratio could not be used as the sample was too small.
D) none of the above.