We know in 450 k array filtering, packages remove probes with >0.01 p.value, but here p.value determines what? Which characteristic of probes define by it ?
I'm assuming that you're referencing detection P values, which represent the confidence that the scanner detected what it's reporting. These are internally determined within the scanner (as far as I know). These p values are useful to remove spurious observations that are close to background, as they're not worth testing, and this is particularly advantageous on methylation arrays where there are hundreds of thousands of tests.