In the present article some robust procedures estimating the mean values and their variances of a set of results are dealt with. Non-parametric methods using quartiles, various kinds of medians, weighted results and dominant cluster are mentioned. For data with significant skewness log-transformation or lambda-transformation are recommended, although the latter is time-consuming and both transformations have influence upon the shape of original data. From practical point of view approaches using quartiles and/or medians are to be preferred because of their simplicity.

        Key words: mean value, variance, robust statistics.