Imputing Values Below Limit of Detection (LOD)

Question

Hi all,

What seems as if a seemingly easy function, isn't available in JMP especially for those involved in forensic chemistry/toxicology. We are often presented with data that is below an instrument threshold (LOD). To compare the results with multivariate statistics we need to impute the values. What this imputed value should be is a point picked out from a distribution in log space since our data is often log-normally distributed. However, JMPs multivariate imputation platform does not allow you to put in a limit of detection before running this imputation. We cannot have values larger than the limit of detection and we cannot have values less than zero.

Does anyone know how to manually code in these LOD into JMP? Can it be done in the missing value menu?

Thanks

Byron_JMP · Answer

Highly recommend taking a look at JMP 16 Pro. There is a new LOD column property so that when you run your regression in the Generalized Regression Platform it automatically deals with left or right or both left and right censored data.

Mark_Bailey · Answer

I am not sure what methods (or platforms) you are using, but some of the estimation platforms properly treat LOD results as left censored data. The Generalized Regression platform in JMP Pro, for example, accepts censored data like this.