In the research on mathematics education, numerous Likert-type instruments estimating attitudes toward mathematics are sometimes composed of factors with a high correlation, which can make it difficult to assign the statements from the scale to each estimated factor. Further, the measurement of attitudes is usually done by adding the scores but ignoring the existence of possible differences in the importance that each item can have in its factor estimation. A revision of the methodology for the study of attitudes toward mathematics for the correct validation of the instrument is carried out in this research, and an estimation of the factors of attitude is proposed using the calculation of the weighted average of scores based on the importance that each item has in the explanation of its factor, which is given by a structural equation model. This methodology has been applied to Auzmendi’s scale of attitudes toward mathematics measurement in a sample of 1293 university students. The factors were estimated using simple and weighted averages; significant differences have been obtained in the measurements, as well on those shown with the organization proposed by Auzmendi.