Tests

What 3 main things decide which inferential test to use?

Whether the study is testing difference or correlation, the level of measurement and the design.

What is a test of difference?

A test used to see whether two conditions or groups differ significantly.

What is a test of correlation?

A test used to see whether two co-variables are significantly related.

When does design matter for inferential tests?

When it is a test of difference.

What design is related?

Repeated measures or matched pairs.

What design is unrelated?

Independent groups.

When are parametric tests used?

Usually with interval data and stricter assumptions.

Which AQA tests are parametric?

Pearson's r, related t-test and unrelated t-test.

Which AQA tests are non-parametric?

Sign, Spearman's rho, Wilcoxon, Mann-Whitney and chi-squared.

What is the golden order for choosing a test?

Difference or correlation -> level of measurement -> related or unrelated if needed.

SNoRD stands for what test?

Sign test.

SNoRD = when do I use it?

Test of difference, nominal data, related design.

What does the R in SNoRD remind me of?

Related design.

What does the D in SNoRD remind me of?

Difference.

COS stands for what test?

Spearman's rho.

COS = when do I use it?

Test of correlation, ordinal data.

Pearson's r abbreviation in your set?

CIP.

CIP = when do I use it?

Test of correlation, interval data, parametric.

WORD stands for what test?

Wilcoxon signed-rank test.

WORD = when do I use it?

Test of difference, ordinal data, related design.

MIDO stands for what test?

Mann-Whitney U.

MIDO = when do I use it?

Test of difference, ordinal data, unrelated design.

DIRRT stands for what test?

Related t-test.

DIRRT = when do I use it?

Test of difference, interval data, related design, parametric.

UNIITD stands for what test?

Unrelated t-test.

UNIITD = when do I use it?

Test of difference, interval data, unrelated design, parametric.

CANDI stands for what test?

Chi-squared test.

CANDI = when do I use it?

Test of association, nominal data.

Which test do I use for nominal + related + difference?

SNoRD.

Which test do I use for ordinal + correlation?

COS.

Which test do I use for interval + correlation?

CIP.

Which test do I use for ordinal + related + difference?

WORD.

Which test do I use for ordinal + unrelated + difference?

MIDO.

Which test do I use for interval + related + difference?

DIRRT.

Which test do I use for interval + unrelated + difference?

UNIITD.

Which test do I use for nominal + association?

CANDI.

What is the sign test used for?

To test for a significant difference in nominal data from a related design.

Why is the sign test non-parametric?

Because it uses nominal data and does not assume normal distribution.

What is Spearman's rho used for?

To test for a significant correlation using ordinal data.

Why is Spearman's rho suitable for ordinal data?

Because it works with ranks.

What is Pearson's r used for?

To test for a significant correlation using interval data.

Why is Pearson's r parametric?

Because it is used with interval data and stronger assumptions.

What is Wilcoxon used for?

To test for a significant difference using ordinal data from a related design.

What is Mann-Whitney U used for?

To test for a significant difference using ordinal data from an unrelated design.

What is a related t-test used for?

To test for a significant difference using interval data from a related design.

What is an unrelated t-test used for?

To test for a significant difference using interval data from an unrelated design.

What is chi-squared used for?

To test for a significant association between two sets of nominal data.

Does chi-squared test difference or association?

Association.

What does correlation mean in inferential testing?

A relationship between co-variables.

What does association mean in chi-squared?

A relationship between categories of nominal data.

For most difference tests like sign, Mann-Whitney and Wilcoxon, what kind of observed value is needed for significance?

Usually the observed value must be equal to or smaller than the critical value.

For tests like chi-squared, t-tests and correlation tests, what kind of observed value is needed for significance?

Usually the observed value must be equal to or greater than the critical value.

Why must I always check the relevant critical values table?

Because significance depends on sample size, tails and the specific test

Nominal + related + difference

SNoRD

Ordinal + correlation

COS

Interval + correlation

CIP

Ordinal + related + difference

WORD

Ordinal + unrelated + difference

MIDO

Interval + related + difference

DIRRT

Interval + unrelated + difference

UNIITD

Nominal + association

CANDI