Modelagem de dados longitudinais complexos no R: desenvolvimento de um pacote estatístico

Abstract

The emergence of digital computers in the second half of the 20st century revolutionized the use of statistics, as it made it possible for methodologies, which were previously unfeasible to be put into practice, to be explored. In particular, longitudinal surveys began to be increasingly used in a context of large surveys. Problems then arose about incorporating the sampling design of the survey into mathematical methodologies and also the translation of such mathematical abstractions into practical computational terms. With regard to fixed and random effects longitudinal models and structured covariance longitudinal models, both involving the complex sampling scheme, the present work aims to present computational techniques capable of solving the second part of the statistical modeling problem highlighted above, through the discussion and creation of a package in the R programming language, called ’clm’. Throughout the study, good programming practices aimed at the development of a clean and sustainable code were discussed, as well as possible strategies to be adopted as the solution for the problem of implementing longitudinal models, such as the use of flowcharts. In the end, the results of such practices were presented analyzing the main functions of the developed package and also an application exemplifying its use, seeking to highlight the benefits provided by this work.