Understanding the Significance of Censored Data

The regulatory world seeks to ascertain if the levels of chemicals in the environment are at a “safe” level. Such an objective leads to dismissing results that are reported as below the detection level (BDL) as being satisfactory. Although this meets the regulatory objective, when attempting to understand the mechanisms that lead to low concentration, such data can be very important. As such, it must be treated in a statistically appropriate manner. The BDL or censored data need to be incorporated into evaluations. Moreover, there is a need to understand the data’s limits of precision. Right censored data are routinely handled for toxicological and reliability studies. BDL values produce left censored values that with a simple transfer can be flipped to right censored values by subtracting all values from a number larger than the data. Then the data can be processed using the statistical tests familiar to toxicologists. Examples of low-level stack emission data and ambient air data are examined in this article. The examples indicate that data should be clearly qualified to ensure that the reader understands the limitations. Using censored data to evaluate factors contributing to measured levels can lead to erroneous conclusions, raise concerns in the public, and lead to potentially unnecessary remedial efforts.