Warning: fopen(/home/virtual/epih/journal/upload/ip_log/ip_log_2024-12.txt): failed to open stream: Permission denied in /home/virtual/lib/view_data.php on line 95 Warning: fwrite() expects parameter 1 to be resource, boolean given in /home/virtual/lib/view_data.php on line 96
There are many epidemiologic studies to find the relationship between disease occurrence and categorized exposure variables which are measured in continuous scales. Recently, it has been found that the differential misclassification can arise when exposure variables are observed with measurement errors and categorized for the analysis. Even though the differential misclassification leads to serious misclassification bias, there is no theoretical attempt to correct the misclassification bias occuring in these circumstances. In this paper, we propose a new statistical method to reduce the misclassification bias due to dichotomizing continuous exposure variables. Since the exposure values are more likely to be misclassified when the true exposure values are close to the cutoff point, the method proposed here gives smaller weights in these case and more weights when these values are far from cutoff point. Simulation studies are performed to compare the bias and the power of the proposed method compared to other methods. Main results are as follows: 1. The proposed method produces the smaller bias and the higher power than the simple method which modifies misclassified data using sensitivity and specificity of exposure misclassification. 2. When the standard deviation of the measurement error are moderately large, the bias and the power of the proposed estimate are somewhat better than those of the modified estimate which excluding the misclassified observations in the analysis. In conclusion, the method proposed here is found to be useful in epidemiologic studies when continuous exposure variables are obtained with measurement error and categorized in the analysis.