Banca de DEFESA: LARISSA MARA COSTA MORAIS

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : LARISSA MARA COSTA MORAIS
DATE: 29/01/2021
TIME: 14:00
LOCAL: https://conferenciaweb.rnp.br/webconf/simone-queiroz-da-silveira-hirashima
TITLE:

ANALYSIS OF STATISTICAL METHODS APPLIED TO THE THERMAL COMFORT INDICES CALIBRATION FOR URBAN ENVIRONMENTS

 

KEY WORDS:

PET; UTCI; Calibration; Statistical methods

PAGES: 100
BIG AREA: Engenharias
AREA: Engenharia Civil
SUBÁREA: Construção Civil
SPECIALTY: Processos Construtivos
SUMMARY:

The calibration of thermal comfort indexes for urban environments is done through the use of statistical methods, which correlate the numerical value of the index with the response on people's perception of the thermal sensation, in order to determine the thermal comfort bands. Studies aimed at calibrating the Universal Thermal Comfort Index (UTCI) and the Equivalent Physiological Temperature (PET) index are increasing in many countries. However, there is currently no standardization of the procedures to be adopted during the calibration process, especially with regard to statistical treatments. In this context, the present study seeks to analyze the statistical methods most used in the calibration of these two indices in order to verify which of the methods most used by the scientific community has the greatest adjustment. A survey of previous studies and verification of the most used statistical methods was carried out. Calibrations were performed using the R software, using three statistical techniques most cited in previous studies and three different scales of thermal perception (3, 5 and 7 points), for a sample of 3,630 subjective responses from previous studies in the city of Belo Horizonte - MG, tropical climate region. Then, a comparison of the results obtained and analysis of the statistical methods was performed. As a result, 34 calibrations were surveyed, involving 50% of the calibrations on the international stage. The PET index and the methods of linear regression analysis, ordinal logistic regression and probit analysis, respectively, achieved a greater number of calibrations. Calibration by linear regression was not viable, as it did not fit the database under study in a linear fashion. The calibrations of the PET and UTCI indexes by the ordinal logistic regression method reached a smaller thermal comfort range compared to the probit analysis for 3, 5 and 7 points of the thermal perception scale. However, it was analyzed that the limits of these thermal comfort bands reached very close values and there was no impact variation. Through the analysis of the model's quality measures (Bootstrapping and AIC) combined with the number of points on the thermal perception scale, it is proved that the ordinal logistic regression for 3 categories, both for the PET index and for UTCI, reached a greater adjustment in this study.

BANKING MEMBERS:
Presidente - SIMONE QUEIROZ DA SILVEIRA HIRASHIMA
Interno - CONRADO DE SOUZA RODRIGUES
Externo à Instituição - ALOISIO JOAQUIM FREITAS RIBEIRO - UFMG
Externo à Instituição - EDUARDO L KRÜGER - UTFPR