Naive statistical analyses for COVID-19: application to data from Brazil and Italy

Abstract

This article is a direct consequence of the authors' desire to discuss the role of statistics in data analysis. The analysis of coronavirus (COVID-19) databases are used as to show simple, but powerful statistical frameworks. We do believe that models for assessing future trends in temporal data in general, and in cases and/or deaths of COVID-19, belongs to the area of (Bio)Statistics. Just as engineers use knowledge of physics, chemistry and often architecture, when constructing bridges, buildings and roads, statisticians use knowledge of mathematics, computer science and even physics for modelling, analysing, and forecasting in order to transform data into information. While the statistician's contribution is rarely acknowledged, everyone knows that a building is a work of an engineer. Nonetheless, nowadays statistics has been gaining the attention that it deserves due to the rise of big data and data science that was built on the foundations of statistics. This article shows that, even with only basic knowledge of statistics, one can adequately collaborate with the community in dealing with very important issues such as the COVID-19 numbers. In order to model and to obtain predictions we use well-known distributions to statisticians working on survival analysis: gamma, Weibull and log-normal distributions. We also make use of singular spectrum analysis, a simple non-parametric time series methodology, for an analogous purpose. Survival analysis is a research area widely used in Biostatistics and even in Reliability, while time series analysis is widely used across areas where the data is measured along the time.