Measurement Error Model And Medical Application

Measurement errors often lie in medical researches. People can reduce thequantity of measurement errors by means of quality control in some researches.But as things stand, measurement error is too great to ignore in other researches.If researchers lose sight of this so great measurement error, their results tend tobe unreasonable. In this article, the author discusses the basic theory of measurement errormodel. Measurement errors always do attenuation on regression, and reliabilityratio is the indicator of this attenuation. Measurement errors make ordinary leastsquare invalid. Computer simulation proves that measurement error model ispreponderant if there are errors in variables. There are two plots in the article torepresent the attenuation of measurement error and the relationship betweenmeasurement error model and orthogonal regression. The identification ofmeasurement error model is somewhat complex. Error variances are necessaryto estimate parameters in measurement error model. An important alternativemethod is to adopt instrumental variable. In this research, 318 doctors in a hospital were sampled randomly. Theoccupational stress inventory revise edition (OSI-R) was used in the survey. Inorder to discover the relationship between occupation stress and occupation task,individual coping style, the author constructed measurement error models byusing instrumental variables. The results show that in medical affairs occupationtask is important to occupation stress and rising to the occasion can reduceoccupation stress to a certain degree. Reliability ratios indicate how greatmeasurement errors are. Measurement error models are fit to the data, so theresults are credible. Measurement error models are based on the classical models, and bringmeasurement errors in models. Measurement error model discover the effects oftrue values by removing the effects of measurement errors. When researchersfind that measurement errors are too great to neglect in their surveys, theyshould no longer fasten their old models and measurement error models areadvisable alternative.