4.1.1.1. Suicide data

The data for all suicides (I, II) in Finland over the 16-year study period of 1980-95 were obtained from official statistics of the Finnish Statistics Centre. The data included the monthly frequencies of suicides classified by year of occurrence, gender (male, female), method of suicide (drowning, gas, hanging, jumping from a height, poisoning, shooting, wrist-cutting, other methods) and age group (≤39, 40-64, ≥65 years) of a suicide victim. In subsequent analyses suicide methods were categorised as violent (hanging, drowning, shooting, wrist-cutting, jumping from a height) and non-violent (poisoning, gas, other methods).

4.1.1.2. Population data

The annual mean population values were extracted from the Yearbook of Official Statistics of Finland (Statistics Finland, 1985 and 1995) and used in calculations of annual violent and non-violent suicides per 100 000 mean population.

4.1.1.3. Data for peak-to-trough differences

In order to compare seasonal swings in suicides between studies from countries of different geographical distances from the Equator, a review was performed (II). All suicide seasonality articles published during 1966-96 were identified through a Medline-database search (keywords ”suicide” and ”season”). Only original articles, which actually reported the monthly frequencies of suicides, were included in the present study.

The following variables were extracted: origin (country) of a study sample, distance from the Equator as measured from the approximate midpoint of the country (km), total number of suicides, monthly number of suicides, length of the follow-up period (years), months of peak and trough incidences of suicides, and statistical significance for overall seasonality.

4.1.2.1. Seasonal variation

The seasonal variation of suicides was examined by months and seasons (I, II). The winter season was defined as December, January, and February; the spring season as March, April, and May; summer as June, July and August; and autumn as September, October and November.

The seasonality of suicides (I, II) was assessed with the method described by Wonnacott and Wonnacott (1990). Firstly, the ordinary chi-square test for multinomials was used as an overall measure of deviation. The null hypothesis stated that suicides within a time interval (month, season) occur with a probability proportional to the length of that time interval.

Secondly, to locate more precisely a time interval in which there was a possible departure from the null hypothesis, a ratio and its 95% confidence interval for each time interval were calculated according to:

where π is the observed proportion of suicides, π ₀ is the expected proportion of suicides when the null hypothesis is true, and n is the total number of suicides in a study population. If the null hypothesis was true, the value of the ratio π /π ₀ had to be approximately 1.00. The null hypothesis was rejected if the 95% confidence interval of the ratio did not include the value 1.00. The calendar effect due to unequal numbers of days in time intervals (months, or seasons) and the effect of leap years were taken into account, when calculating the expected frequencies of suicides.

4.1.2.2. Subgroup analyses

The differences in monthly proportions of suicides were analysed according to gender and the age group of a suicide victim. The ordinary method of comparing the differences in two large sample proportions was chosen as a way to assess the statistical significance of differences (Wonnacott & Wonnacott 1990).

4.1.2.3. Peak-to-trough differences

The peak-to-trough difference was calculated by subtracting the ratio of trough month from the ratio of peak month. With this statistical approach it was possible to estimate the magnitude of the seasonal swing of suicides (I, II).

4.1.2.4. Time trends

Time trends in the rates of violent and non-violent suicide (II) over the years 1980-95 were assessed graphically and with an ordinary linear regression analysis. In the regression analysis, the dependent variable was the rate of suicides per 100 000 mean population and the independent variable was the year of occurrence, recoded as being time = 1 at 1980, 2 at 1981, 3 at 1983, etc., and 16 at 1995.

Time trends in the seasonal variation of violent and non-violent suicides (II) were examined separately in three non-overlapping time periods (1980-84, 1985-89, 1990-95). The statistical methods used were the chi-square test for multinomials and the ratio statistic and its 95% confidence interval (see chapter 4.1.2.1.). Furthermore, in each time period, the months for maximum and minimum ratios of suicides were identified and the peak-to-trough differences were calculated in order to find out, whether there have been any changes in these variables over a time course.

4.1.2.5. Statistical software

The statistical software SPSS 6.1 for Windows (Norusis 1995) was used in all statistical analyses.