According to statistician's views, randomization methods are considered most reliable for detecting cloud seeding effects. But evaluating the results of an operational cloud seeding program are rather diffcult. Brief summaries of defects and experience of weather modification were confined to main weather modification activities that appear to be based on acceptable physical principles and which have been tested in the fields. An argument about two randomization experiments which won high praise is discussed. By citing some examples in point, the presently recommended analyses for evaluating weather madification experiments are based on distribution free inference techniques.