1. Chi-squared Flashcards
Tests for Correlation and Association??
If we want to look for correlations, we use Pearson’s r or Spearman’s rho.
If we’re counting observations in categories, we use Chi-squared.
Describe the conditions for when to use Chi-squared (or Fisher’s exact).
> Counting observations; testing for association
Two independent groups
Categorical variables (not scale, scales require a test of correlation)
Non-parametric
What are linear relationships?
Variables and constant are connect; change in one -> change in other variable.
These can be positive (both increase together), or they can be negative (one increases, the other decreases).
What tests do we use for linear relationships (para and non)?
Parametric data ==> Pearson’s (product-moment) correlation coefficient (R or r).
Non-parametric ==> Spearman’ Rank order correlation coefficient or Spearman’s Rho or p.
What can tells us how strong a linear correlation is?
Correlation coefficient (r value)
> perfect corr = 1 (+/-)
No corr = 0
When discussing relationships of data, what must you report?
> The correlations co-efficient (r-value)
> The p-value
> The number of pairs of data (n)
What refers to the effect sizes for correlations?
List the values and meanings.
The effect size for correlations is the r-value squared (r^2).
Large 0.5 and above
Moderate 0.3 to 0.49
Small 0.1 to 0.29
Ignore Below 0.1
If we aint got fisher’s exact, what do we look for?
> 2x2 crosstabs –> SPSS gives Fisher’s exact
> 3x2 table –> SPSS doesn’t give Fisher’s exact, but we need to look at Pearson’s χ^2 value and the p-value.
What do we report with Fisher’s exact?
Only report the p-value.
What do we report with Chi-square?
We report the degree of freedom, the critical value, and the p-value.
Conditions for the McNemar test
> This is for within subject design.
> The McNemar test assesses whether the change is due to chance, or whether there is a trend in one direction.
Chi squared single sample
Sample is compared to population distributions (actual vs expected)
