Hello!

Good question indeed. You may find answers in this paper:

Article A Bayesian Multiple Imputation Method for Handling Longitudi...

When calculating ratios with a Limit of Detection (LOD) in the denominator, there are a few different approaches that can be used, depending on the specific context and the desired level of accuracy. Here are a few examples:

Hello dears,

I don't know if there are other similar/same questions, but in any case, it's good to open a discussion on this topic!

In my particular situation:

Jonathan Maurer

Maybe I do not have so much data (N=3) for statistical estimations. What's your opinion?

Musa Titus Sesay

So I'm going to use the conservative approach. It should be more reasonable in my case.

Thanks for the valuable advice and for sharing the insightful references

You are welcome Piergiorgio Cianciullo

Piergiorgio Cianciullo

In my (very limited) personnal experience, I don't really like LoD. Because there is no accepted standard procedure for LoD measurement, it might be quite misleading.

In the peptides analysis field for example, the guidelines do not ask for a determination of LoD to validate a quantification method, but rather use the LLoQ (see CLSI 2021 guidelines for example). Then we estimate values

Thanks Jonathan Maurer !

There are some good recommendations here. However, my caveat is that it exemplifies the danger of trusting (or using) statistics as a substitute for valid science. If you can't OBSERVE it in some way (

By the way: the determination of LOD as the result of a lab procedure follows a statistical approach!

Steven Cooke

I totally agree with you philosophically.

Just to be clear, in my practical case, I have to confront concentrations of heavy metals in exposed samples with background unexposed concentrations. I found 3 to 4 orders of magnitude higher than LOD in exposed samples.

Now I try to follow your statement: it is true that there is a difference (

For interpretation of your data, you certainly want to set your results in relation to some interpretative cassifications! There are many, e.g., LD50, or WHO standard permissible limits of heavy metals, etc.

Concerning LOD: next year you buy a new machine and LOD is two orders of magnitude lower! Your comparison will be pretty questionable by then. Thats why I propose to report true concentrations as well.

Piergiorgio Cianciullo , yes, this discussion is good. My point is that 'forcing' an answer (or making one up) to 'fit' a particular ratio intended for other analyses is not the best approach to solving that particular problem. LOD is a real thing, and a moving target. But sometimes we misunderstand what "0" means, or whether adding decimals to a number is a practical outcome. Maybe a different measure or comparison than the existing ratio is required.

Medical instrument usually have detection limits. which are corelated to medical terms. below detection limit means , safe value but it depends on various other factors.