research methods: Choosing a statistical test

Question 1

What should we consider when choosing a stat test?

Answer 1

1. Are we looking for a difference or a correlation?
2. What experimental design was used?
3. What level of measurement was used?

Question 2

What does a difference vs correlation refer to?

Answer 2

• Relates to the aim or purpose of the investigation
• Will be evident in the wording of the hypothesis

Question 3

What experimental designs are there and what data do they produce?

Answer 3

• Independent groups -> unrelated
• Repeated measures -> related
• Matched pairs -> related

Question 4

What levels of measurement are there?

Answer 4

• Nominal data
• Ordinal data
• Interval data

Question 5

What is nominal data?

Answer 5

• Data that is represented in the form of categories
• Is discrete because one item can only appear in one category

Question 6

What is ordinal data?

Answer 6

• Data that is ordered in some way
• Does not have equal intervals between each unit
• Lacks precision because it is based on subjective opinion rather than objective measurement
• Raw data is converted into ranks e.g. 1st, 2nd, 3rd and these ranks are used in the testing

Question 7

What is interval data?

Answer 7

• Data that is based on numerical scales that include units of equal, precisely defined size -> a specific scientific statistic/measurement e.g. heart rate
• the most precise and sophisticated form of data in psychology

Question 8

What are the parametric tests?

Answer 8

• Related t-test
• Unrelated t-test
• Pearson’s R

Question 9

What are the advantageous features of parametric tests?

Answer 9

• Much more powerful and robust than other tests
• Use only the most trustworthy data
• Only use interval data
• If a psychologist can use a parametric test, they will, as these tests may be able to detect significance within some data sets that non-parametric tests cannot

Question 10

What are the three criteria of parametric tests?

Answer 10

• Data must be interval level because parametric tests use the actual scores rather than rank
• Data must be drawn from a population that would be expected to have a normal distribution