In this thesis, corrosion data were subjected to statistical analysis for the purpose of obtaining meaningful estimates of the specific rate of reaction and the energy of activation and obtaining a response equation for prediction of weight gain. A two-stage analysis was used in the estimation of specific rate and activation energy. First, an estimate of the reaction rate for each test run was obtained from linear regression analyses of weight gain on time. A weighted linear regression was then performed with reciprocals of estimated variances of rates as weights; the estimated order of reaction was obtained. It was shown that the mechanisms of the two possible reactions, i.e., formation of aluminum oxide or AlC13x6H20, are different. At the lower HCl partial pressures (Al2O₃ the product}, the specific rates and the activation energy were estimated in two ways: (1) the method of weighted linear regression when stoichiometric first order is assumed; (2} the method 0f weighted covariance analysis when empirical order is assumed. The data in the A1c13x6H20 portion were inconclusive. The prediction equation was obtained by an analysis of covariance; the model was that of a two-way classification with time as a covariate. Main effects ( temperature and pressure) were estimated. Finally, a response equation, tor prediction of the weight of corrosion product (A12O₃ ), was found. This equation is valid for HCl partial pressures o,QO) to 2.90 mm Mg; temperatures 17 °C to 44 °C, time of exposure $0 to 130 minutes. / Master of Science