Chapter9 Chi-square

9.0.1 Overview

107 cases were gathered wherein a former Hogwarts student was a defendant in a magic court case.
Below is a table outlining the number of students who were convicted or acquitted in their case along with the house they belonged to.


9.0.2 Observed Data

Gryffindor Hufflepuff Ravenclaw Slytherin Total
Convicted 6 9 5 23 43
Acquitted 25 13 21 5 64
Total 31 22 26 28 107

We need to conduct a chi-square test to determine whether a students’ former house has any association with whether they were convicted or acquitted.
To do so, we must first calculate the number of convicted/acquitted students expected in each house if the null hypothesis is assumed true.


9.0.3 Expected Data

Gryffindor Hufflepuff Ravenclaw Slytherin
Convicted
Acquitted

For each cell, calculate:

\[ E = \frac{(\text{row total}) \times (\text{column total})}{\text{total N}} \]


9.0.4 Residuals

Gryffindor Hufflepuff Ravenclaw Slytherin
Convicted
Acquitted

Now we need to measure the discrepancy between the observed and expected data for each cell with:

\[ \frac{(O - E)^2}{E} \]

In other words, square the difference between each observed and expected frequency and divide that number by the expected frequency for that cell.


9.0.5 Statistical Test

A \(\chi^2(df = \_\_\_) = \_\_\_\)

Insert the correct degrees of freedom in the parentheses:

\[ df = (r - 1)(c - 1) \]

Set \(\chi^2\) equal to the sum of the residuals.
The “critical value” for this test statistic is approximately 7.825.
Would you end up rejecting the null hypothesis?


9.0.6 Standardized Residuals

Gryffindor Hufflepuff Ravenclaw Slytherin
Convicted
Acquitted

Calculate the standardized residuals for each cell:

\[ \text{Standardized Residual} = \frac{(O - E)}{\sqrt{E}} \]

Circle any standardized residuals greater than 3.


9.0.7 Notes

  • Be sure to show all intermediate steps in your calculations.
  • Interpret your findings: does house membership appear to be associated with court outcomes?