Comparing means and proportions II

Question 1

What is the analogue of the one sample t-test?

Answer 1

One sample Wilcoxon signed rank test

Question 2

When would we use the Wilcoxon signed rank test?

Answer 2

With ordinal, discrete or skewed continuous data

Question 3

What are the assumptions of the Wilcoxon signed rank test?

Answer 3

• The observations are randomly and independently drawn

* At least ordinal data (interval data, or ratio variables)

Question 4

When reporting the results of the Wilcoxon signed rank test what do we report?

Answer 4

We double click the box
We report whether or not the null hypothesis was rejected
We reported the standardised test statistic and the asymptotic significance (p-value)

Question 5

What is the analogue of the two independent samples t-test?

Answer 5

Mann – Whitney U test aka Wilcoxon rank sum test

Question 6

What are the assumptions of the Mann Whitney U test?

Answer 6

Two independent samples
• The pairs of observations are randomly and independently drawn
• At least ordinal data (interval data, or ratio variables)

Question 7

When reporting the results of the Mann Whitney U test what do we report?

Answer 7

We double click the box
We report whether or not the null hypothesis was rejected
We reported the Mann Whitney U value and the asymptotic significance (p-value)

Question 8

What is the analogue for the paired samples t-test?

Answer 8

Paired sample Wilcoxon signed rank test

Question 9

What does the paired sample data need to be, to be able to conduct the paired sample Wilcoxon signed rank test?

Answer 9

Each sample needs to be asymmetrical, BUT the difference between the two MUST be symmetrical. There is currently no test for paired data where the difference is also asymmetrical/skewed

Question 10

When reporting the results of the paired Wilcoxon signed rank test what do we report?

Answer 10

We double click the box
We report whether or not the null hypothesis was rejected
We reported the standardised test statistic and the asymptotic significance (p-value)

Question 11

What are the assumptions of the paired sample Wilcoxon signed rank test?

Answer 11

• The pairs of observations are randomly and independently drawn
• At least ordinal data (interval data, or ratio variables)
• The (paired) difference is a symmetrical variable

Question 12

Non-parametric tests are tests of ….

Name the 3 non-parametric tests

Answer 12

medians
Wilcoxon signed rank test
Mann Whitney U test (aka Wilcoxon sum rank test)
Paired sample Wilcoxon signed rank test

Question 13

What type of sample is better suited for non-parametric tests?

Answer 13

Skewed data
Smaller sample sizes (less than 30) - However, this is no longer such an important assumption due to computer programs such as SPSS

Question 14

Why are non-parametric tests worse than parametric tests?

Answer 14

They are not as powerful

Question 15

When would we conduct exact tests?

Answer 15

When the assumptions of the categorical data tests are not met.
When the below assumptions are violated
• Up to 20% cells can have expected count less than 5. - WHEN OVER 20% OF CELLS HAVE AN EXPECTED COUNT LESS THAN 5
• The minimum expected count is larger than 1 - WHEN THE MINIMUN EXPECTED COUNT IS LESS THAN 1

Question 16

When the assumptions are violated on the one sample chi squared test, what exact test do we use?

Answer 16

one sample exact test

Question 17

When we conduct the one sample exact test what do we report?

Answer 17

We report the exact significance (exact p value)

Question 18

When we have two independent samples that violate the assumptions of the pearsons chi squared test, what test do we use instead?

Answer 18

Fisher’s exact test

Question 19

When we conduct the fishers exact test what do we report?

Answer 19

The fishers exact test significance value (fishers exact p value= )

Question 20

What do we do if our categorical independent samples test involves more than 2x2 data?

Answer 20

Then need to selected manually on SPSS

Question 21

What happens if our McNemar test is violated as there is less than 25 discordent cells?

Answer 21

It doesnt matter. The McNemar p-value is always valid, as long as we have paired categorical data. SPSS instead of printing the asymptotic p-value prints the exact test. However, SPSS always prints the exact p-value (binomial), even if the assumptions hold!

Question 22

When we conduct the McNemar test on data that does not meet the assumptions what do we report?

Answer 22

McNemar Browker value, df and p value