Two statistical methods for the analysis of data on the evolution of the chemical composition of cold snow (<0Β°C) in the field (Lac Laflamme, Quebec) were compared. The methods used on the data were regression analysis (One sample per sampling date over a long cold period) and ANOVA (replicate samples on a restricted number of sampling dates over shorter periods). The relative power of the tests to determine the detectable amplitude of chemical changes was derived from the theoretical power of the tests under comparable conditions of sampling (number of observations) and from the estimated error variances of the measured data.The results of the study on the evolution of sulfates (SO4) concentrations in discretely identified snow strate clearly showed that for six of the eight strata, significant losses of SO4 occurred in snow during cold periods. The relative amplitude of the significant losses varied between 1% per day and 4% per day depending on the initial concentrations in the snow and the prevailing meteorological conditions.The analysis of the data also demonstrated that for the same number of samples, the regression analysis is more efficient in detecting the chemical changes in snow than the alternative ANOVA method. The use of this information to plan sampling programs of cold snow under both field and laboratory conditions is discussed.