The estimation of fitnesses from population data.

HIS paper is concerned with certain statistical problems which arise in conTnection with attempts to estimate fitnesses from the genotype frequency data produced by artificial populations in which rapid changes in genotype frequencies are taking place. The special situation of concern here is the one in which the investigator wishes to obtain several estimates of fitness, each taken from small segments of the selection curve, in an attempt to evaluate patterns of change in fitness as the experiment proceeds. Examples of such attempts are DOBZHANSKY and LEVENE (1951), POLIVANOV (1964), TOBARI and KOJIMA (1967) and YARBROUGH and KOJIMA (1967). The problems of estimation arise when the genotypes are classified and counted in each generation at a stage when the selection process is incomplete. PROUT (1965) pointed out that if selection is incomplete at the time of observation, then estimates of net fitness could not be obtained by methods then in use. The present paper constitutes an extension of this analysis.