Statistical Analysis and Relevance

Air sampling and industrial hygiene engineering.

Martha J. Boss & Dennis W. Day

CHAPTER 4

Statistical Analysis and Relevance

This chapter contains theoretical and real-world discussions about statistical analysis. It gives special emphasis to situations where knowledge of statistical relevance is a prerequisite for sampling adequacy. It also illustrates the difference between log-normal and normal distribution and para­metric monitoring data.

4.1 DEFINITIONS

In statistical analysis and relevance, certain standard definitions are used. The follow­ing standard definitions and examples illustrate the basic concepts.

• Accuracy: Agreement of the measured value (i.e., empirical value) and the "true
value" (i.e., accepted reference value) of the sample given valid sampling tech­
niques, proper sample preparation, and reliable and accurate instrumentation.
and/or other procedures.

Accuracy is often estimated by adding (or "spiking") known amounts of the target parameters. For asbestos quality assurance (QA) sampling, accuracy is evaluated by comparing analyses of duplicate samples that have been evaluated in proficiency in analytical testing (PAT) round robins (for air samples) or NIST National Voluntary Laboratory Accreditation Program (NVLAP) proficiency test­ing (for bulk and transmission electron microscopy [ТЕМ]).

Accuracy is a measure of the bias of the method and may be expressed as the difference between two values, a ratio, or the percentage difference.

• Analysis: Combination of sample preparation and evaluation.

• Audit: Systematic determination of the function or activity quality.

• Bias: Systemic error either inherent in the method or caused by measurement sys­
tem artifacts or idiosyncrasies.

• Blind sample: Presented to the laboratory as indistinguishable from field sam­
ples (syn: performance audit samples). All field blanks are to be presented to the
laboratory as blind samples for asbestos air samples.

• Calibration: Comparative procedure in which singular measurements are evalu­
ated against an accepted group of measurements. The evaluation may be against
a primary, intermediate, or secondary standard.

• Calibration curve: Range over which measurement can take place (syn: standard
curve, multipoint calibration).

• Calibration standard: Instruments or other measurement techniques used to
evaluate the measurement system. Ideally, these standards do not directly incor­
porate or use the target parameters to be measured.

• Chain-of-custody: Defined sample custody procedures that must be followed to
document the transition from field collection to subsequent transfer sites (com­
mon carriers, laboratories, storage facilities, etc.).

• Check standard: Used to verify that the initial standard or calibration curve
remains in effect. It ideally incorporates standard materials (syn: daily standard,
calibration check or standard, reference standard, control standard, single point
response factor, single point drift check).

• Comparability: Confidence with which one set of empirical data can be com­
pared to another.

• Completeness: Amount of valid data obtained from a measurement system
compared to the amount expected.

• Data quality: Totality of data parameters that identify ability to satisfy or repre­
sent a given condition; includes accuracy, precision, representativeness, and com­
parability.

• Data reduction: Using standard curves to interpret raw data.

• Data validation: Review process that compares a body of data against a set of cri­
teria to provide data adequacy assurance given the data's intended use; includes
data editing, screening, checking, auditing, verification, certification, and review.

• Detection limit: Minimum target parameter quantity that can be identified, i.e.,
distinguished from background or "zero" signal.

• Double-blind sample: When neither the composition nor identification of the
sample is known to the analyst.

• Duplicate sample (field or laboratory): Sample divided into two portions, with
both portions carried through the sample preparation process at the same time.
For asbestos air samples, field duplicates are air samples collected at the same
time as the compliance air samples, and lab duplicates are portions of one filter
that are fixed and analyzed separately.

• Environmentally related measurements: Field or laboratory investigations that
generate data involving chemical, physical, or biological parameters characteris­
tic of the environment.

• Field blanks: Generated at the time of sampling, field blanks provide a check on
contamination, starting with the sampling process and proceeding through the full
analysis scheme. For asbestos fiber concentration sampling, field blanks are filter
cassettes transported to the site and exposed to ambient conditions. The filter caps
are removed from the filter cassettes; however, a vacuum air pump is not used to
pull air across the filter cassettes. Thus, the cassettes are exposed to the environ­
mental airstream of the surrounding environment outside the asbestos control area.

• Good laboratory practices: Performing a basic laboratory operation or activity so
as not to influence data generation quality.

• Instrument blank: Used to obtain information on instrument aberration
absence/presence. The measurement instrument is presented with materials
normally within the instrument and cycled through the measurement sequence.
The resulting signal is then defined as the baseline instrument signal level.

• Internal standard: A nontarget parameter added to samples just prior to measure­
ment to monitor variation in sample introduction and stability and to normalize

data for quantitation purposes. Internal standards are not usually used in bulk (phase light microscopy [PLM]) or phase contrast microscopy (PCM); however, these standards may be applicable to ТЕМ protocol.

Laboratory blank: Prepared in the laboratory after receipt of samples from the field. These blanks are prepared using a material assumed not to contain the tar­get parameter. Lab blanks for asbestos sampling are filter membranes obtained from filter cassettes that have been retained in the laboratory without removal of the filter cassette caps. The lab blank is a check on all the chemicals and reagents used in the method as well as the influence of the general laboratory environment (syn: analytical blank, system blank, method blank, process blank). Measurement: Creating quantitative data from a prepared sample. Method check sample: Prepared in the laboratory by spiking a clean reference matrix with known quantities of the target parameters. For asbestos air-sampling analysis, method check samples are previously prepared filters evaluated by sep­arate analysts within the same laboratory. These duplicate analyses are defined in the National Institute of Occupational Safety & Health (NIOSH) 7400 method as quality assurance, and the acceptable statistical parameters are outlined therein. Method detection limit: Minimum quantity that a method (i.e., both sample preparation and target parameter measurement steps) can be expected to distin­guish from background or "zero" signal. This limit takes into account losses dur­ing preparation and measurement and instrument sensitivity that may contribute to qualification or quantification of results. This limit does not apply to physical parameters (i.e., density, temperature).

Performance evaluation (PE) sample: Sample with known "true" values that is presented to the laboratory as a "performance evaluation sample." These sam­ples are biased by the analyst's knowledge of the intent of the sample. For asbestos air-sampling analysis, "true" value samples may be defined as the PAT samples with their inclusive statistical ranges.

Precision: Measure of the reproducibility of a set of results obtained under simi­lar conditions. Precision is determined by multiplicate analysis of samples, dupli­cates, replicates, or splits. Standard deviation is used as a measure of precision. Procedure: Systematic instructions and operations for using a method of sam­pling or measurement.

Proficiency sample: Samples for which known composition values are available for accuracy comparisons. The composition values may be qualitative, quantita­tive, or statistical ranges of acceptable qualitative/quantitative results. Quality assurance: An orderly assemblage of management policies, objectives, principles, and general procedures by which a laboratory outlines the methods used to produce quality data. QA is an intralaboratory function. Note: The NIOSH 7400 method defines QA in terms of both intralaboratory and interlabo-ratory methods and/or sequencing. However, for the purposes of specified QA/quality control (QC) documents, inter laboratory methods are defined as QC. Quality control: Routine application of procedures used to develop prescribed performance standards in the monitoring and measurement of standards. QC is an interlaboratory function.

QC samples: Analyzed concurrently with field samples to insure that analytical systems are operating properly, i.e., in control. These samples provide an estimate of the precision and accuracy of the sampling and analysis system. QC samples for asbestos sampling are sent between laboratories for interlaboratory compar­isons of methodology and analytical proficiency.

• Quality of method: Degree to which the method functions free of systemic error,
bias, and random error.

• Quantitation limits: Maximum and minimum levels or quantities of a reliably
quantified target parameter. These limits are bounded by the standard curve lim­
its and are generally related to standard curve data.

• Reagent blank: Used to identify contamination sources. These blanks incorpo­
rate specific reagents during sample preparation to identify lab blank contami­
nate sources (syn: dilution blank).

• Recovery/percent recovery: Generally used to report accuracy based on the
measurement of target parameters, comparison of these concentrations, and cor­
relating these measurements to the predicted amounts. Recovery in asbestos air
sampling is the statistical percentage differential observed during accuracy eval­
uation (i.e., percentage difference in fiber concentrations).

• Replicate samples (field or laboratory): A sample is divided into two portions
and is processed as two completely separate and nonparallel samples, i.e., pre­
pared and analyzed at different times or by different people. Field replicates in
asbestos air sampling are defined as filters that are obtained from two separate
filter cassettes drawn separately or from a y-juncture. These are then transported,
fixed, and analyzed separately. Lab replicates are taken from a singular filter,
which is sectioned, fixed, and analyzed separately.

• Representativeness: Degree to which data accurately and precisely represent a
parameter variation characteristic at a sampling point and portray an environ­
mental condition.

• Sample custody: Verification and documentation procedure for the transfer of
samples from the field to the laboratory, within the laboratory, and to the final
storage or disposal destination.

• Sample Operation Procedure (SOP): Procedure adopted for repetitive use when
performing a specific measurement or sampling operation.

• Sample preparation: Transformation of the sample into appropriate forms for
transfer and/or measurement.

• Sampling: Removal of a process stream representative portion or a portion of a
larger quantity of material for subsequent evaluation.

• Sensitivity: An instrument's detection limit given the minimum quantity of a tar­
get parameter that can be consistently identified, i.e., distinguished from back­
ground or "zero" signal by the instrument; ideally established using the materials
that are used for standardization.

• Split samples (field or laboratory): A sample, divided into aliquots, that is sent
to a different laboratory for preparation and measurement. These split samples
may be replicates or duplicates that are then defined as splits when sent to
another laboratory for QC analysis. For asbestos air samples this shipment
involves either the shipping of capped filter cassettes (field split) or the shipping
of fixed slides (lab split).

• Standard materials: Materials, such as mixtures «of the target parameters at
known concentration and purity, used to carry out standardizations. For asbestos
air sampling these materials may be PAT samples.

• Standardization: Establishing a quantitative relationship between known target
parameters input and instrument readout.

• Target parameters: Entity for which qualitative or quantitative information is
desired.

• Trip blanks: Essentially field blanks that do not have the caps removed. These
blanks provide insight into the contamination generated as a result of the shipping
process. These blanks are generally not required for asbestos sample shipment.

4.2 EXAMPLE—OUTLINE OF BULK SAMPLING QA/QC PROCEDURE

Bulk sample analysis procedures are defined in the NIOSH method for PLM and the NIOSH 7402 method for ТЕМ. Because bulk sample analysis is done by an independent laboratory off-site, this QA/QC document will not address bulk sampling as a field-verified procedure. The NVLAP is currently used to ascertain laboratory effectiveness. The contractor must provide proof that NVLAP certification is current for the laboratory des­ignated to receive both the initial bulk samples and the 10% bulk sample duplicates.

The optical properties and ID of fibers are as follows:

• Determine 13 specific items for asbestos.

• ID other fibers with some optical data.

• ID matrix components.

In addition record

• Special procedures or solvents

• Sampling of layers

• Deviations from EPA test procedure

The 10% bulk sample duplicates are provided by physically dividing 10% of the sam­ples collected. This division should occur concurrently with the collection of field samples. The rationale for splitting the samples at a time and location removed from the field col­lection site must be provided.

All field collection procedures, sample labeling, and transport and disposal procedures must be addressed in the QA/QC document. Provide the rationale for the classification and remediation of errors. The following is a sample of error classification:

Major Error

• False positive

• False negative (asbestos actually >1%)

• Incorrect asbestos type classification

• Analysis quantification in error by 25%

Minor Error

• Incorrect ID of tremolite or actinolite as another type of asbestos

• False negative (asbestos actually >1% or trace)

• Analysis quantification in error by 15%

• Incomplete lab data sheets (repeated omissions may equal major error)

Дата добавления: 2015-11-16