A Simple Method for Testing Monthly Statistical Rates for Normalization
Till Roenneberg · Chronobiology International · 1990
The spectral analysis of birth statistics, expressed in monthly rates, shows a strong periodic component even if a single child is born every day of the year. This is, of course, due to the different number of days per month. Most human or animal statistics as well as agricultural and biometeorological variables are only available as monthly, and not daily rates or values. If spectral components of time series are of interest, it is necessary to normalize monthly rates. This necessity has been a pre-analytical routine for quite some time, and various methods of normalization are described in the literature (1,3,7): the rates are either expressed as the daily mean within a month or as rates of normalized months with 31, 30 or 30.4 days (the latter results in a 365-day year). Each of the different methods leads to the same result, if the analysis can be based on relative deviations from the mean.In the majority of the statistical material, the chosen method of normalization is clearly stated. But if a study tries to cover many different statistical populations and as many years as possible, the statistical material is often based on several (partly secondary) sources and photocopied extracts from large volumes. This can lead to confusion whether the used statistics contain normalized or crude original monthly rates.