"Evaluation of Similarity Between Variables by their Native Values and by Proportional Deviations from their Own Exponential Trendlines (pd) -with pd Calculators, Regressions and Visual Analyses - Examples of CHD Subgroups in Different Periods Between 1951-87"

Abstract

Native' data of Finnish coronary heart disease mortality (CHD) of rural (rur) and urban (urb) subgroups available only by 3-year means, 3ym, from period 1951-87, (for calculation: 1952-88) revealed, that male (M) (M.CHD.rur) in period 1952-77 associated positively with smoking, consumption of alcohol and sugar, but negatively with milk fats), oppositely to (female) F.CHD.rur and F.CHD.urb and mainly oppositely with M.CHD.urb and pig microangiopathy (MAP) (mortality). Numerical evaluation was challenging, especially because CHD's showed visual similarities. For explaining such a situation is earlier presented a method using "proportional deviations of parameters from their own exponential trendlines" (pd) between start point (α) and end point (ω). Pd data, as well as 'native' data, are here analyzed through Pearson correlations, regressions and visual charts by three periods 1952-86, 1952-77 and 1963-86 (3ym). A ready calculator is presented for producing pd data.Results: Periodical Pearson correlations between M.CHD and F.CHD data of middle-aged people in rural regions with 'native ' [and pd] data were in 1952-86 +0.22 [+0.91], 1952-77 -0.32 [+0.95] and 1963-86 +0.91 [+0.82]. Periodical means +0.27 [+0.89], SD +0.50 [+0.05].Associations between M.CHD.rur and F.CHD.rur were inverse in period 1952-77, opposite to other periods. 'Native' data showed highest reciprocal association in period 1963-86 (+0.91), pd data in 1952-77 (+0.95). Summary:Pd data [based on proportional deviations from exponential trendlines between start (α) and end points (ω)], via their Pearson correlations, regressions and visual charts can give a new (?) method to evaluate similarity, especially simultaneousness in variation and possibly "pick up" some details, undetectable by linear regression analysis with native data.