Sequences of samples of time-varying quantities processed by an algorithm are burdened by errors that vary randomly in successive samples (type A) and errors that have constant values (type B). These errors propagate differently through the algorithm, and therefore their specifics must be taken into account in the process of determining the uncertainty of the result at the output of the algorithm. The article describes a mathematical apparatus that allows the description of the result in the form of a measurand interval formed on the basis of a probabilistic description of both types of error. Analytical considerations are illustrated with examples.