Heteroskcedasticity Two-way Linear Regression And Using To Test Concrete Strength Of Building Structure

When the in-site data is limited, it will be an urgent problem to be solved, that is, how to set up a better precise empirical formula. In former researches, two-way linear regression method is offered under such a condition that borrows the widening sample size of lab data or historical data, and it is an avail exploration for above mentioned problem. In practical projects it is difficult to meet the hypothesis that the in-site data has the same variance as the quoted data, so this method has some limitations in practice. After further study, for the problem of heteroskedasticity, a two-way linear regression method is offered on the condition of heteroskedasticity, and the situations of the standard deviation relationship (the ratio of the standard deviation of in-site data and the standard deviation of borrowing data) of both known and unknown are discussed respectively, and corresponding point estimation and corresponding interval estimation are given.On the condition that the standard deviation is given, the estimated value of empirical formula given by heteroskedasticity two-way linear regression has the nature that it is unbiasedness and has the minimal variance. On the condition that the standard deviation is not given, it is testified by a large quantity of Monte Carlo that this method is not only advanced than the traditional methods, but also much more approaching the true value of matrix comparing with two-way linear regression method.If the standard deviation is known, interval estimation method of empirical formula is given in this dissertation; if the standard deviation is not completely known, approximate evaluation method of empirical formula in confidence interval is given.If apply heteroskedasticity two-way linear regression to test intensity of structural concrete, although in-site data and history or lab data have different deviations, the problem can be solved efficiently that in-site data is less so can not establish the higher precise in-site forced test curve. At the same time, this method also can be applied into the similar situations of other fields.