Calculations for the BeLPT cont'd

DOE-SPEC-1142-2001

B.8.2 The LAV method was selected because its use of medians resists the influence that

outliers can have on estimates calculated from a small number of replicates. If

outliers are not considered to be a problem, then conventional and more efficient

least squares estimates (i.e., using means instead of medians in B.2 and the usual

estimate of the standard deviation in B.4) could be used. Also, note the sqrt(pi/2) is

not included in StErr[Ln(SI)].

B.8.3 An alternative outlier resistant method that calculates estimates of the Ln(SI)s and

their standard errors could also be used. An example of another outlier resistant

method is given in EHP96 (see Quasi-likelihood estimation). Alternative

approaches to robust regression are described in Chapter 25 of "Applied Regression

Analysis," Third Edition (see Reference B.11.1).

B.8.4 The only requirement if another method is used is that appropriate values of the

Ln(SI)s (B.3) and their standard errors (B.5) shall be used in B.6 to obtain the

values of the standardized Ln(SI) for use in 10.3.2 to determine if two or more

beryllium treated wells show a positive response.

B.9 The results of all of the calculations and additional statistics are combined into a single

laboratory LAV report (see Table B-IV).

Note that Sm and the residuals in the top panel of Table B-IV have been multiplied by 100,

i.e., they are in Ln percent units. The estimates of Sm for the control wells and treated wells

for Day 5 and Day 7 with and without "pooling" are provided in the last panel of Table B-

IV. An "overall" pooled estimate is also provided. These values are used to calculate the

standard error[Ln(SI)] and to evaluate the amount of variation within control and treated

groups on Day 5 and 7.

The LAV report is used for quality control and to help in the interpretation of BeLPTs that

are not confirmed abnormals or normals. The BeLPT in Table B-IV has two large

standardized Ln(SI) s (D5be10 and D5be100) indicating a positive test (see Sec 9.2.2).The

standardized maximum SI for this test, (see Sec B10) is:

standardized maximum SI = (0.98 - M )/ SD = (0.98 - 0.081)/0.34 = 2.64.

This is below the cut point of 3.1 (see Sec 9.2.1) for the reference data set. The values of

Sm are high on Day 7 and the standardized Ln(SI) for D7be10 is -3.98, indicating cell

killing in at least one well. This test was interpreted as "borderline" by the laboratory, and

subsequent follow-up found this patient to be SENSITIZED.