A very good question indeed. And I apologize for my little outburst. Wasn't aimed at anyone in particl'r.

A model or analysis which aligns with other data is more likely to be "true" than one which continuously runs into contradictions or misalignments of various sorts compared to other data.

A piece of information is as useful as it seems to impart value to other information. Newton's laws are --by this test--much more important and valuable than Sarah Palin's shopping list.

Monotheism as espoused by, say, the Baptist Church could seem to be terribly valuable because it offers one explanation for everything. Trouble is, there is too much other information that doesn't line up, such as the fossil record, or the strange condensation of time and hashed-up sequence of events in Genesis, for a couple of examples.

There are a lot of other indexes you can use, as well; they can serve to improve the certainty with which you hold a given datum or set of data, but in the final analysis, data is never absolute. So you do the best you can with gradual approximations.

A