next »

[joeroberts]

I'm working with some contingency tables and there appears to be some confusion about the residuals.  Agresti (2007) refers to standardized Pearson residuals which result from dividing the difference between the observed and expected results by the standard error.  According to him, any value above 2 or 3 means the independence model is a poor fit for that cell.  By contrast, SPSS calls them standardized adjusted residuals.   And Andy Field calls them standardized residuals and says that any value above 1.96 can be seen as statistically significant because this is a z-score and 1.96 is the cut-off value for a .05 level of confidence.
Can anyone clarify this?