There has been little agreement as to whether reproduction or similar demog
raphic events occur seasonally and, especially, whether there is any univer
sal seasonal pattern. One reason is that the seasonal pattern may vary in d
ifferent populations and at different times. Another reason is that differe
nt statistical methods have been used. Every statistical model is based on
certain assumed conditions and hence is designed to identify specific compo
nents of the seasonal pattern. Therefore, the statistical method applied sh
ould be chosen with due consideration. In this study we present, develop, a
nd compare different statistical methods for the study of seasonal variatio
n. Furthermore, we stress that the methods are applicable for the analysis
of many kinds of demographic data. The first approaches in the literature w
ere based on monthly frequencies, on the simple sine curve, and on the appr
oximation that the months are of equal length. Later, "the population at ri
sk" and the fact that the months have different lengths were considered. Un
der these later assumptions the targets of the statistical analyses are the
rates. In this study we present and generalize the earlier models. Further
more, we use trigonometric regression methods. The trigonometric regression
model in its simplest form corresponds to the sine curve. We compare the r
egression methods with the earlier models and reanalyze some data. Our resu
lts show that models for rates eliminate the disturbing effects of the vary
ing length of the months, including the effect of leap years, and of the se
asonal pattern of the population at risk. Therefore, they give the purest a
nalysis of the seasonal pattern of the demographic data in question, e.g.,
rates of general births, twin maternities, neural tube defects, and mortali
ty. Our main finding is that the trigonometric regression methods are more
flexible and easier to handle than the earlier methods, particularly when t
he data differ from the simple sine curve.