Abstract
It is shown that the freeze distribution is a mixture of the distribution of freeze-date and the simple dichotomous distribution of freeze and freezeless years. This is applied both nonparametrically and assuming a normal distribution of freeze date to three stations at three different thresholds to obtain the probabilities of freeze before or after any date. The distribution of the freeze-free period is developed and application made to one of the stations to obtain probabilities of the freeze-free period being less than a given time interval. The expressions for the mean freeze-date and freeze-free period are also developed and estimates made for the stations treated.
