Re: Entering '<1' and 'BDL'

On Tue, 20 Apr 2010, Tom Keller wrote:

> I was thinking of something like a UV absorbance of a sample too dilute to
> accurately measure.

Tom,

Our current client has this situation: the amount of SO2 (a surrogate for
Hg) coming out the roaster stacks is so low that the calibration is at the
low end of the measuring instrument. Therefore, calibrations fail on a
frequent basis. However, the EPA defines the scale to be used so they're
caught in a bind.

> You would get a reading of 0.04 or something, but since your
> spectrophotometer is only accurate between say, 0.4 - 3.0, it's
> essentially noise. Nonetheless, I can tell that by any value below my
> detection limit, so the value, though not accurate, would distinguish
> between not measured and measured but below the detection limit.

So, if you have results you know are below 0.4, what do you enter as a
value in the database? And, if the meter pegs at 3.0 and is trying to bend
around the pin, what value to you enter into the table for that reading? The
answers have to be understood by the techs who will be entering the data via
a Web browser form.

> Similarly, I should think, if you measured something beyond, or off-scale,
> you could tell that from the value entered too. So for the
> spectrophotometer example, below 0.4 is noise and above 3.0 is off-scale.
> So you don't know what their accurate values are, you know they were "too
> low" to detect, or "too high" to measure with your instrument.

Gimme' a f'instance of how you would record these readings in a table
where that attribute is of type NUMERIC. That's what we're looking for. :-)
We need a way to record such instances that make sense, are consistent and
easy for techs and mill workers to use, don't confuse data analyses (such as
averages and trends), and would be acceptable to regulators.

Rich