Two multiple linear regression (MLR) models were developed with the aim to estimate the decomposition temperature of a series of polyimides. Two parameters, Tni and Tai, corresponding to the temperature of 10 % weight loss of the sample, determined by dynamic thermogravimetric analysis under conditions of N2 inert atmosphere and air, respectively, were used as a criterion for thermal stability. The obtained MLR models correlate thermostability with a series of characteristics of the studied polymers, such as Van der Waals volume, density, molecular weight, number of aromatic cycles, number of C=O bonds, number of CH3 groups and the number of CF3 groups. The results showed that the MLR models can be successfully used to predict the thermal stability of polyimides, the mean percentage errors being below 3%, regardless of the work environment.