CMStatistics 2018: Start Registration
View Submission - CMStatistics
B1393
Title: Analysis of trends in congenital anomalies prevalence: Method comparison simulation study Authors:  Jan Klaschka - Institute of Computer Science of the Czech Academy of Sciences (Czech Republic) [presenting]
Marek Maly - Institute of Computer Science of the Czech Academy of Sciences (Czech Republic)
Antonin Sipek - Thomayer Hospital Prague (Czech Republic)
Abstract: Data of several national health registries covering years 1994 - 2015 are analyzed in order to detect trends in the prevalence of different kinds of congenital anomalies (birth defects). The standard analysis tool in such a situation is the Poisson regression, but non-statisticians frequently use the simple linear regression instead. A provocative question then arises, whether - in our specific conditions - the former approach is better enough than the latter one. In order to answer, a simulation study was designed, comparing the two methods both under the null hypothesis of no trend, and in presence of trends of different shapes and intensities. The settings examined mimicked the reality by using the actual series length and yearly numbers of newborns, and realistic range of event frequencies. The results speak uniformly in favour of the Poisson regression: The simplistic approach (linear regression) was outperformed in the terms of test power over a broad spectrum of parameter values used.