Percentage deviation and standardized residual are both measures of the degree to which an observed chi-square cell frequency differs from the value that would be expected on the basis of the null hypothesis.Q
For each cell, percentage deviation is calculated asQ
xxx
observed — expectedexpected
x 100
Thus, a percentage deviation of +15% within a cell indicates that the observed frequency is 15% greater than the expected, while a percentage deviation of -15% indicates that the observed frequency is 15% smaller than the expected.
In the special case of df=1, the calculation of percentage deviation incorporates a correction for continuity:Q
xxx
|observed — expected| —0.5expected
x 100
The resulting value is then given a positive sign if observed>expected and a negative sign if observed<expected.Q
The standardized residual for a cell in a chi-square table is a version of the standard normal deviate, z, calculated asQ
xxx
z =
observed — expectedsqrt[expected]
In the special case of df=1, the calculation of the standardized residual incorporates a correction for continuity:Q
xxx
z =
|observed — expected| —0.5sqrt[expected]
The resulting value of z is then given a positive sign if observed>expected and a negative sign if observed<expected.
The chi-square value that results from a chi-square analysis is equal to the sum of the the squares of the standardized residuals.
Assuming the null hypothesis to be true, and providing that the expected value for a cell is at least 5, values of the standardized residual belong to a normally distributed sampling distribution with a mean of zero and a standard deviation of ±1.0.