Local polynomial estimation of TGA derivatives using logistic regression for pilot bandwidth selection

A method is proposed to find a function that fits traces resulting from the decomposition of polymers during a thermogravimetric analysis (TGA) experiment. The aim of this work is to find an expression that relates the weight of the sample with time (or temperature) and to study certain features like derivatives and critical points. It has been possible to keep in the asymptotic trend of the TGA trace by using a local logistic type regression. The typical asymptotic tendency of the weight loss is perfectly reproduced.