O wyznaczaniu kwantyli rozkładu w szeregach rozdzielczych przedziałowych
Streszczenie
Interval frequency distributions used as a secondary data do not provide full information
about the distribution of the variable of interest. In order to estimate quantiles, it
is common to use one of interpolation formulae based on the assumption that the distribution
of the variable is uniformly distributed in each interval. This assumption, if not
met, may result in large bias. Author suggests using different formulae, which do not require uniform distribution
in intervals, and depend on the frequencies of neighbouring intervals. Simulation experiments
were applied for the normal and lognormal distributions to assess efficiency of
both kinds of interpolation formulae. Different population sizes and different numbers of
intervals in the frequency distribution were also considered.