GVM valuation model considers that PBR is the function of ROE. There are two coefficients in the function. One of them is value coefficient (k), the other is growth coefficient (m). The main purpose of this study is to improve the GVM model by regulating coefficients k and m of the GVM model by using the industry, scale and the risk, which would result in more accurate prediction of PBR. In this research, to calculate the value of k and m, we categorized the data of listed companies from 2005 to 2016 into four data sets according to industrial types (Construction, Finance, Traditional, and Electronic industry). Also, based on the scale or risk, we divided them into ten data sets from low to high. To overcome measurement error in the independent variable(ROE) of GVM model, adopted the Errors-in-variables Regression(EIV) to precisely estimate coefficient k and m. The result shows that: (1) There are some significant biases in the regression curve generated by traditional regression analysis, but it can be correct by using the errors-in-variables regression analysis. (2) Industry category affects the value of k and m. Construction industry has the highest value of k, followed by finance, and the traditional and electronic industry have the lowest value. In contrast, the effects of industry category on the value of m appear the opposite order. (3) The scale of the company has the most significant effects to the value of k and m. The greater the scale is, the less the value coefficient k is whereas the higher the growth coefficient m is, which indicates that large companies get the higher persistence of earning. (4) The risk of the companies won’t significantly affect the value of k and m.