Design and Analysis of Gauge R&R Studies: Making Decisions with Confidence Intervals in Random and Mixed ANOVA Models

Consider a balanced random model with Q sources of variation represented by the ANOVA shown in Table 8.1. Assuming all random errors are independent and normally distributed, the
are jointly independent and each
has a chi-squared distribution with n q degrees of freedom for q = 1 , , Q (see, e.g., Hocking [31, p. 460]). These results are used to construct confidence intervals for all four parameters shown in Table 8.2, where c q, d r, and e t ? 0. Note that ? 3 and ? 4 differ in two respects. First, ? 3 has nonnegative linear combinations in both the numerator and the denominator, whereas ? 4 allows negative linear combinations in the numerator. Second, the same expected mean square cannot appear in both the numerator and the denominator of ? 3, but it can in ? 4. Although ? 3 is a special case of ? 4, it is useful to consider the two cases separately. Note that neither ? 3 nor ? 4 allows negative coefficients in the denominator.
| Source of variation | Degrees of freedom | Mean square | Expected mean square |
|---|---|---|---|
| Factor 1 | n 1 | | ? 1 |
| Factor 2 | n 2 | | ? 2 |
| | | | |
| Replicates | n Q | | ? Q |