If you can assure that the properties with which the data should be verified, are verified with antelation and during the selection of them, and they have been obtained with precise measure, you could say that your data are collected validly and reliably.
It can be possible by doing crosschecking of the data and results with existing model. For example, Suppose assuming your data following normal distribution, then by checking the residual and graph fitting of the probability distribution function (PDF) it can be checked the validity of data and their reliability too.