(no subject)

William Tivol tivol at news.wadsworth.org
Tue Dec 19 16:53:05 EST 1995

Flemming Vester (fvest at mi.aau.dk) wrote:
: What is the right way to do an MPN (most probable number)analysis?

: There seem to be some confusion in the literature on how the analysis should be
: done. I have come acros two different ways:

: 1)
: Some researchers make one single dilution serie, and transfer samples from this
: serie into f.ex. three replicates.

: 2)
: Other researchers for example, do three individual dilution series.

: As far as I can see, the two used methods are different, and require two
: different statistical methods.

Dear Flemming,
	Yes, you are correct, but, if the errors in making the dilutions are
much smaller than the other errors, then the two methods are equivalent.
Looking at the statistics, for method #1, each of the replicates is the same
dilution, so an unknown factor (nearly = 1) must be incorporated into each of
the three replicates, whose variation from one another results from the other
possible errors.  Sometimes there is an assay which will give, e.g., a least-
squares fit to the unknown factors, but if the other errors are large, this
fit will be unreliable (possibly more so than the dilution process).  For me-
thod #2, the variation of the samples from each dilution incorporates the
error of the dilution along with the other errors, but if the other errors
are large, the contribution of the dilution error is negligable.  Method #1
separates the dilution errors from the others; whereas, method #2 does not.
Doing the assays both ways will tell you if the dilution errors are impor-
tant--I suspect they are not.  Good luck.
				Bill Tivol

More information about the Bioforum mailing list