Statistics 5

Question 1

What is the most common type of data analysis in geography?

Answer 1

Non-parametric data

Question 2

What is a benefit and a problem with using non-parametric data analysis?

Answer 2

Problem: only simple arithmetic (non-normal)
Benefit: less assumptions made which makes it easier to work with

Question 3

What are inferential statistics?

Answer 3

Tests of difference

Question 4

What does a one-way chi-squared test do? What is the parametric equivalent?

Answer 4

Test for the difference between sample and population for nominal and ordinal data. Parametric equivalent is a one-sample t-test

Question 5

What are we interested in from what SPSS produces for a one-way chi-squared test?

Answer 5

The value produced (not that interesting) and the p-value (very interesting) because this is the significance value

Question 6

How do you interpret the p value SPSS produces for a one-way chi-squared test?

Answer 6

Normally

Question 7

What does a two-way chi-squared test do? What is the parametric equivalent?

Answer 7

Tests for difference between two samples for nominal and ordinal data. Parametric equivalent is the a two-sample t-test

Question 8

What is cross tabulation and what does it enable us to do?

Answer 8

Cross tabulation enables us to compare two different responses in a survey

Question 9

What is an example of cross-tabulation?

Answer 9

If one question on a survey asks for Gender then as well as this being data collected about the people we can also use it as a lens from which to analyse responses to other questions such as alcohol consumption. We can then see how the alcohol consumption differs along gender lines.

Question 10

How do we interpret the p-value that SPSS produces for a two-way chi-squared test?

Answer 10

normally

Question 11

When is a two-way chi-squared test invalid?

Answer 11

If more than 20% of the expected values for each result/type of answer in a survey have a count less than 5

Question 12

Give an example of when a two-way chi-squared test is invalid

Answer 12

If when looking at opinions on gay marriage where the responses that people can provide are between strongly disagree, disagree, neutral, agree, strongly agree (5 values for each of the 5 categories of response) then if the expected result for 1 (20%) of these categories is below 5 (i.e. E(X) is 4) then the test is invalid

Question 13

What do statisticians commonly do to solve the invalid two-way chi-squared problem?

Answer 13

they combine categories - i.e. if you combine the expected values for two categories then you automatically dramatically increase the chances that the new category that is created as a result of their combination will have an E(X) >5.

Question 14

What have you got to be careful of when solving the invalid two-way chi-squared problem?

Answer 14

You do not simply combine categories that are opposites, i.e. you want to combine ones that are very similar for example “agree” and “strongly agree”

Question 15

What is a Mann-Whitney U test?

Answer 15

This test compares the mean ranks of equal/unequal sample sizes for two samples

Question 16

Explain how a mann-whitney U test works?

Answer 16

It ranks all the responses in their order and determines for each sample what rank the calculated mean value would be within this order. It then compares the ranks of these means for both samples against each other to determine the difference between them

Question 17

What is the Mann-Whitney U test similar to and how is it different?

Answer 17

Two-way chi-squared test for ordinal and ratio and interval data that are not normally distributed.

Question 18

Why is nominal data not able to be tested through the Mann-Whitney U test?

Answer 18

Because it is not as measurable

Question 19

How do you interpret the p-value that SPSS produces for the Mann-Whitney U test?

Answer 19

normally

Question 20

What is a Kruskal-Wallis test and what is the parametric equivalent?

Answer 20

Tests for difference between 3 or more samples. Parametric equivalent is ANOVA

Question 21

Explain how the Kruskal-Wallis test works?

Answer 21

Identical process to Mann-Whitney U test - It ranks all the responses in their order and determines for each sample what rank the calculated mean value would be within this order. It then compares the ranks of these means for each sample with the other samples to determine the difference between them

Question 22

How do we interpret the p-value that SPSS produces for a Kruskal Wallis Test?

Answer 22

normally

Question 23

If we are comparing a lot of samples in a Kruskal Wallis test then what might me want to do to determine the specific relationship between two of them?

Answer 23

Isolate the two samples and carry out a Mann-Whitney U test not a two-way chi-squared

Question 24

What data type are relational statistics not suitable for? What data types are they suitable for?

Answer 24

Nominal not suitable. Ordinal and Ratio and Interval that are not normally distributed

Question 25

What are relational statistics?

Answer 25

Test to determine correlation

Question 26

When carrying out relational statistics what is it typically best to represent ordinal statistics using?

Answer 26

Table

Question 27

How do we interpret the table that SPSS produces for Spearman’s rank?

Answer 27

Cover 1/2 of the table along a line form the top left to the bottom right as the values are the same. Each row has three sub-rows which are correlation coefficient, significance value and N value

Question 28

How do we interpret the significance for a spearman’s rank test in SPSS?

Answer 28

Look at the number of asterisk that are superscript on the correlation coefficient value

Question 29

What does two asterisk mean?

Answer 29

99% confident that it did not occur due to chance i.e. 99% statistical significance threshold.