STATISTICAL METHOD FOR DETERMINING THE GLASS TRANSITION TEMPERATURE FROM DILATOMETRIC DATA.

An objective procedure for estimating the glass transition temperature (T//g) from dilatometric data is described. The method uses the technique of fitting a segmented linear regression model by least squares. The regression model employed may be specified so as to allow a transition either of the first order in the thermodynamic sense or may be constrained to fit a second order. Methods are given for finding statistical confidence intervals for the estimated glass transition temperature (T//g). Experimental data obtained from PET (polyethylene terephthalate) fiber are used for illustration; these data indicate a preference for the second-order transition model. 8 refs.