Desenvolvimento e validação de softwares para análise estatística em Ciências da Saúde

Abstract

The use of statistical tools is of fundamental importance for the validation and association of events in different areas of knowledge. In the health area, it is possible to verify the efficiency of treatments and procedures. However, machine use is limited because there are few unintuitive interfaces, high costs of available programs, and users' difficulty in choosing the appropriate analysis. From this perspective, it is opportune to create more intuitive software to assist users' decision-making, ensuring the data's quality. Given the existing gap, the present study aimed to develop tools for statistical analyses in Health Sciences. For this means, two statistical software were developed. One of them, entitled Statistical Oracle, solves problems related to hypothesis tests. The other one, PPGCS Sample Size, allows calculating the minimum sample size required for studies involving means and proportion. The methodology for software development followed the lifecycle approach, Rational Unified Process (RUP), and prototyping. The requirements were documented with the notation proposed by the Unified Modeling Language (UML). The Statistical Oracle was developed with C# language, while the PPGCS Sample Size was developed in the PHP (Hypertext Preprocessor) language - visual layer HTML 5 (Hypertext Markup Language). The two software scans were tested by comparing the results of the analyses by the tools proposed with those obtained by other softwares, such as SPSS, R, and Excel - 100% agreement was observed. In a usability test compared to the PSPP (Free Software Foundation), the users of the Statistical Oracle were able to solve the proposed problem with a 46% lower time compared to the PSPP. In addition, the success rate with the Statistical Oracle was 82% against 12% for the PSPP, in addition to also receiving a better score in the categories: learning, efficiency, safety and satisfaction. The programs present intuitive interfaces for users; besides, it has educational windows that justify the use of a recommended test and its references. It is essential to highlight that Oracle has quality control of the analyses that may influence the researcher's decision-making. That prevents the user from making wrong choices in the analysis when it is not following the recommended design for the sample. For access and dissemination, programs are available at the PPGCS website at the http://www.ppgcs.unimontes.br . It is concluded that the tools created are reliable for the analyses, easy to handle and practical.