Hi Statalisters,

In clinical medicine, it's very common that the machine has detection threshold. For example, the detection threshold is 100 for one indicator , the value of the indicator that machine reports could be like this:

HTML Code:
321.24
320.26
298.54
254.56
216.87
180.65
156.47
123.15
118.46
105.23
<100
The percentage of the value "<100" is not very low to be ignored (as missing).

One solution is: I can transform the continuous variable to a categorical variable.

Any better ideals?

Thank you for your time!