Measurement Errors

The measures play an important role in statistics. We don't have the access to the true values, therefore, we will need the estimation on the true value. We can measure the variables we are able to observe with error. In literature, we model/predict the ideal outcomes based on the observed measure with a normally distributed error.

The measurement with error usually includes the error from the device, the error from the adherence to the device, the error from the observer.

Each type of device has its own limitation. They may not possible target the true value, instead it may get the true value plus or minus some tolerate errors. This sometimes called the systematic error. It could hardly being avoided, unless the technology being improved.

The error from the behavior of using the device inappropriately can usually being avoidable. It can be verified by validity the measures and by taking average of several observations. However, sometimes it is hard to always get the valid measure, depending on the nature of the data. For example the non-adherence to the device may be common in some study, and will result the missing values mixed with the error in measurements. In this situation, it could be hard to differentiate the measures.

The error from the observer, is frequently studied by discover the demographic nature of the observer. Usually the conclusion could be, certain kind of observers may have large amount of error. This can be done by comparing the errors contributed by different types of observers.

How the study can be designed to valid the measures and distinguish between the types of measurement is really need some investigations. Usually will need the design nature of the study to take this into account, other than at the end of the study to discuss about the issue. The investigators will need to take more effort on the measurement errors before collecting the data.