Skip to main content

Table 7 Blank correction

From: Wrangling environmental exposure data: guidance for getting the best information from your laboratory measurements

Approach (see Additional file 4 for example of this approach with real data):
1. Which blanks to use?
 □ If detects are spread across all types of blanks (e.g., field, solvent method, matrix), we use all blanks for blank correction. Otherwise we use field blanks. We try to keep our blank correction approach consistent with our MRL approach.
2. Which chemicals get corrected?
 □ If > 5 blanks:
   For each chemical, we use a one-sided one sample sign test (special case of binomial test with p = 0.5) to determine whether the median of blanks is statistically significantly different from zero. True and estimated detects are treated as positive values and non-detects as negative values.
   ▪ We blank-correct chemicals with a sign test p-value < 0.05.
   ▪ However, if the number of blanks is relatively small (10 or fewer) we consider blank correction even when the sign test does not produce a significant result. The sign test does not take into account the magnitude of the levels detected in the blanks nor does it distinguish different types of blanks (i.e., field and lab).
    • For example, if we have 3 field blanks and 4 lab blanks, and we see consistent levels detected across all field blanks and all but one lab blank, we would consider blank correcting even though the sign test would produce p > 0.05.
 □ If ≤ 5 blanks (i.e., for a small dataset):
   With five or fewer blanks, the sign test will never be significant. In this case, we blank-correct chemicals with 100% detects in blanks.
3. Blank correction:
 □ Calculate the median value of the blanks, with non-detects set to ½ lab’s reporting limit and using all values (i.e., estimated and true detects).
   It is useful to pause here and assess the value being used for blank correction. Is it based on an estimated value below the MRL? What will be the percent change in the median, comparing the original to the blank-corrected data?
 □ Subtract median blank value from all sample results.
 □ Subtract median blank value from the MRL (determined as in Table 6).
 □ We are explicit about whatever procedure we use to decide whether or not to perform blank correction (sign test or other) and about the statistic (e.g., median, mean) and amount used for correction.
 □ Any presentation of measurements (e.g., summary statistics) should use blank-corrected values because they may be compared with measurements in other studies.
 □ For statistical analyses such as regression and correlation performed within the dataset, non-blank-corrected data can be used.