Critical values for the Grubbs test
Testing one suspect value: if G = |suspect − x̄|/s exceeds the critical value, the value is an outlier at that confidence level.
| 3 | 1.154 | 1.155 |
|---|---|---|
| 4 | 1.481 | 1.496 |
| 5 | 1.715 | 1.764 |
| 6 | 1.887 | 1.973 |
| 7 | 2.020 | 2.139 |
| 8 | 2.127 | 2.274 |
| 9 | 2.215 | 2.387 |
| 10 | 2.290 | 2.482 |
| 11 | 2.355 | 2.564 |
| 12 | 2.412 | 2.636 |
| 13 | 2.462 | 2.699 |
| 14 | 2.507 | 2.755 |
| 15 | 2.548 | 2.806 |
| 16 | 2.586 | 2.852 |
| 17 | 2.620 | 2.894 |
| 18 | 2.652 | 2.932 |
| 19 | 2.681 | 2.968 |
| 20 | 2.708 | 3.001 |
| 21 | 2.734 | 3.031 |
| 22 | 2.758 | 3.060 |
| 23 | 2.780 | 3.087 |
| 24 | 2.802 | 3.112 |
| 25 | 2.822 | 3.135 |
| 26 | 2.841 | 3.158 |
| 27 | 2.859 | 3.179 |
| 28 | 2.876 | 3.199 |
| 29 | 2.893 | 3.218 |
| 30 | 2.908 | 3.236 |
| 35 | 2.978 | 3.316 |
| 40 | 3.036 | 3.381 |
| 50 | 3.128 | 3.482 |
Computed from G = (n−1)/√n · √(t²/(n−2+t²)) with t the upper α/(2n) quantile of the t distribution (ν = n − 2); agrees with ISO 5725-2 Table 5 to the third decimal. Test only the single most extreme value at a time, and look for an experimental cause before rejecting it.