Statistics Meeting Book (July 18, 2018)

RECOMMENDED CONFIDENCE INTERVALS FOR dPOD AND dLPOD

As in TR342, the current recommended confidence interval for dPOD of a single laboratory is that given by eqs.(3) and (4).

The current recommended confidence interval for dLPOD is provided by the following steps:

1. Compute dLPOD from eq.(5). 2. Compute SE(dLPOD) from eq.(6). 3. If the estimate of SE(dLPOD) is greater than zero, then use the t 0.975

or z 0.975

interval of

eq.(7). 4. Otherwise, estimate a nonzero SE(dLPOD) from

+ 2) 2 + SE(dPOD) 2

SE(dLPOD) =

√ { [ 1/(n harm

] / L }

(8)

pooled

where n harm

is the harmonic mean of the individual number of replicates

1 / n harm

=

1 / n 1

+ 1 / n 2

+ … + 1 / n L

(9)

and

2 ] / Σ (n k

= √ { [ Σ (n k

SE(dPOD) pooled

-1) SE(dPOD) k

-1) }

(10)

and use the z 0.975

interval of eq.(7).

4