[Corribus]

Suppose you are making a measurement of concentration and your detection limit is 1 ppb.

You measure four replicates, and the values in ppb are 1.5, 2, 2.5 and <MDL.  What is the appropriate way to report your average value?

My feeling is that you could report a possible range, since the <MDL value is between 0 and 1. So the average must lie between 1.5 and 1.75 ppb. You could apply a potential error (STD) to each limit of the range as well. But I've asked around and gotten a lot of blank looks and conflicting answers. Someone people have suggested you can't report any value at all.  Does anyone have an opinion on this?

[Furanone]

I know the lab I used to work at many years ago would have taken average of the four results with the one <MDL result being assigned a value of 0.

[Arkcon]

I know the lab I used to work at many years ago would have taken average of the four results with the one <MDL result being assigned a value of 0.

Thank you for this, that does seem to be the best approach.  By definition, a result less than the MDL has already been defined as statistically indistinguishable from noise.  So what would you do if there was nothing detected in one repetition?  Same thing, really.

Now I'm curious, what do you do if one rep is below the LOQ?  Likewise call it zero?  Probably not, so long as LOQ is greater than LOD.

[Corribus]

Ok, so you call it zero. You're not the only person who recommended this to me. It might be the most straightforward thing to do.

[Furanone]

If only one rep was done, then it would simply be listed as <LOQ, which would imply there was some of that substance there (since above LOD), but it is going to define it other than it is very small (ie. somewhere between 0 and whatever LOQ for that assay happens to be)