A stochastic framework is presented to enhance the design of textile antennas, by addressing the impact of production uncertainties, substrate bending and compression on the performance of textile antennas. Each effect is considered separately and the probability density functions corresponding to the design parameters undergoing variations are experimentally estimated. In order to accurately quantify the statistical distribution of the antenna's figures of merit corresponding to these variations, we introduce a stochastic collocation method based on a generalized polynomial chaos technique or on Padé approximants.