0000033 OHIO ENVIRONMENTAL PROTECTION AGENCY DIVISION OF EMERGENCY AND REMEDIAL RESPONSE FINAL HOW CLEAN IS CLEAN POLICY July 26, 1991 I. INTRODUCTION The Ohio Environmental Protection Agency (Ohio EPA), Division of Emergency and Remedial Response (DERR) is responsible for the discovery, investigation, enforcement and remediation of unregulated hazardous waste sites. The Division conducts these activities under the authority vested in the Director of the Ohio EPA by the Ohio Revised Code (ORC) Sections 3734 and 6111. The DERR conducts activities at sites where the Division has reason to believe a release or the potential for a release of a hazardous waste, as defined under ORC 3734.01 (J), has or may occur due to the treatment, storage or disposal of such waste. These sites may pose a threat or potential threat to human health or safety or contribute or threaten to contribute to air, water or soil contamination. The DERR generally restricts its activities to sites where the treatment, storage or disposal of hazardous waste occured prior to the enactment of the Resource Conservation and Recovery Act of 1976 (RCRA). Many of these sites are abandoned or uncontrolled and are not currently operating facilities. Therefore, these sites are referred to as "unregulated" hazardous waste sites and are not regulated under RCRA. The DERR works in conjunction with other Federal and Ohio EPA programs in the investigation and remediation of unregulated hazardous waste sites. Activities conducted by DERR are consistent with all applicable state and federal laws (i.e. Clean Water Act, Resource Conservation and Recovery Act, Comprehensive Environmental Response, Compensation and Liability Act). The Division's responsibilities are administered through three major program areas: Emergency Response - Response to chemical and petroleum releases, spills and waste dumping incidents that present or nay present an immediate threat to human health or the environment.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
0000033
OHIO ENVIRONMENTAL PROTECTION AGENCYDIVISION OF EMERGENCY AND REMEDIAL RESPONSE
FINALHOW CLEAN IS CLEAN POLICY
July 26, 1991
I. INTRODUCTION
The Ohio Environmental Protection Agency (Ohio EPA), Division ofEmergency and Remedial Response (DERR) is responsible for thediscovery, investigation, enforcement and remediation ofunregulated hazardous waste sites. The Division conducts theseactivities under the authority vested in the Director of the OhioEPA by the Ohio Revised Code (ORC) Sections 3734 and 6111.The DERR conducts activities at sites where the Division hasreason to believe a release or the potential for a release of ahazardous waste, as defined under ORC 3734.01 (J), has or mayoccur due to the treatment, storage or disposal of such waste.These sites may pose a threat or potential threat to human healthor safety or contribute or threaten to contribute to air, wateror soil contamination.
The DERR generally restricts its activities to sites where thetreatment, storage or disposal of hazardous waste occured priorto the enactment of the Resource Conservation and Recovery Act of1976 (RCRA). Many of these sites are abandoned or uncontrolledand are not currently operating facilities. Therefore, thesesites are referred to as "unregulated" hazardous waste sites andare not regulated under RCRA.The DERR works in conjunction with other Federal and Ohio EPAprograms in the investigation and remediation of unregulatedhazardous waste sites. Activities conducted by DERR areconsistent with all applicable state and federal laws (i.e. CleanWater Act, Resource Conservation and Recovery Act, ComprehensiveEnvironmental Response, Compensation and Liability Act).
The Division's responsibilities are administered through threemajor program areas:
Emergency Response - Response to chemical and petroleum releases,spills and waste dumping incidents that present or nay present animmediate threat to human health or the environment.
Special Investigations - Investigations into environmental crimeallegations that can result in criminal, civil and/oradministrative enforcement.
Remedial Response - Discovery, investigation and remediation ofunregulated hazardous waste sites through enforcement.
The Division has developed this guidance to provide a generaloverview to hazardous waste site investigations and remediationsconducted by the DERR, Remedial Response Program. The primarypurpose of this guidance is to generate a consistent approach forevaluating site investigations and clean up. General proceduresare outlined which can be consistently and uniformly applied toevery situation. This guidance does not present site specificclean-up criteria or guidance. Every site is different in regardto magnitude and nature of contamination, no single approach canbe applied. Therefore, site specific issues will be addressed bythe DERR through individual enforcement agreements.
II. SITE INVESTIGATIONS
The DERR, Remedial Response Program conducts investigations atunregulated hazardous waste sites in a manner not inconsistentwith the National Contingency Plan (NCP), Therefore, the DERRreferences many of the same guidance and criteria documents asthe U.S. EPA Superfund program. Appendix A lists the guidanceand criteria documents that the DERR, Remedial Response Programadheres to in conducting site investigations and clean up. Thispolicy is intended to be used in conjunction with these guidancedocuments.
This section sets forth the general approach that the RemedialResponse Program will take as a site progresses through theinvestigative stage. The purpose of the site investigation is toadequately characterize the nature and extent of contamination.For the purposes of this policy a site is defined as the arealextent of contamination.
A. Determination of Contamination
DERR evaluates sites to determine whether or not there isevidence of a release, or the potential for a release, ofhazardous waste due to past treatment, storage, or disposal. Suchevidence includes environmental field data and/or writtendocumentation obtained through historical records and/orinterviews.
In investigating a site, the DERR considers a site to becontaminated if it meets the following criteria:
* The contaminant(s) detected is a hazardous waste as definedunder the Ohio Revised Code (ORC) 3734.02 (J) an<jf
* Contaminants are present on-site at concentrationssignificantly above background or.
* Contaminants arc present on-site and are not detected inrepresentative background samples.
a. Background
For the purposes of this policy, background is defined ascurrent conditions present at a site and areas surrounding a sitewhich are unaffected by past treatment, storage, or disposal ofhazardous waste. In order to determine the nature and extent ofcontamination, background conditions must be established. Anevaluation of background should be conducted concurrently withthe on-site investigation.
The DERJ* has developed specific guidance which outlines theprocedures to be followed in determining background conditions ata site. The guidance sets forth the DERK's criteria fordetermining the number of samples that should be collected inorder to adequately represent background. This guidance isattached in Appendix B of this document.
b. On-site vs. Background
Once background and on-site conditions have been adequatelycharacterized, results should be evaluated to determine whetheror not contamination exists.
i) Naturally occurring compounds - Many compounds that may bepresent as a result of the treatment, storage or disposal ofhazardous waste may also be present in the environment due to thenatural conditions of the surrounding environment. In order todetermine whether or not these compounds are present due to pastwaste disposal practices, on-site samples must be evaluatedagainst representative background samples.
In such a comparison, if on-site samples exhibit compounds atconcentrations greater than the mean background concentrationplus the product of the tolerance factor and the relativestandard deviation, then contamination is said to exist (SeeBackground Guidance, pg. 18).
If a compound is not detected in background samples then anyfinding in on-site samples greater than the method detectionlimit (MDL) is considered contamination. The MDL is the minimumconcentration of a substance that can be measured and reportedwith 99% confidence that the analyte's concentration is greaterthan zero. MDL's vary from method to method and application ofthe MDL will depend on the appropriate methodology applied.MDL's will be identified for individual contaminants in the sitespecific investigation workplan.
ii) Non-naturally occurring compounds - Compounds that arepresent in the environment due to human activity that aregenerally not naturally occurring must.also be evaluated.For the purposes of a site investigation, one must determine ifnon-naturally .occurring contamination is a result of past
disposal, treatment or storage of waste at the site. Since thesite is defined as the areal extent of contamination theinvestigation must also include areas where waste has beentransported through air, leachate, surface or groundwater. Theinvestigation must also include waste constituents that nay bepresent due to the transformation or biodegredation of hazardouswaste.Background must also be determined for comparison to on-sitesamples, especially if the site is in a highly industrializedarea or where the site is downgradient from another unrelatedhazardous waste source.
If a contaminant is detected at the MDL or greater in on-sitesamples and is not detected in representative background samples,then contamination is said to exist.
If a compound is detected in on-site samples and inrepresentative background samples, then any concentration greaterthan the mean background concentration plus the product of thetolerance factor and the relative standard deviation (seeBackground Guidance, pg. IS) is considered site relatedcontamination and must be addressed. In order to determinecontamination at a site where non-naturally occurring compoundsare found both on-site and in background samples, the DERR mustfirst approve background sampling data and agree that suchcontamination is not related to the site under investigation.
B. Determination of Risk
Once it has been established that contamination exists at a site,it must then be determined whether or not the contaminationpresents a threat to public health or the environment. The DERRconsiders a threat to be present under either of the followingconditions:
* Contamination exists at levels that present or have thepotential to present a current or potential future unacceptablerisk to human health.
* Contamination exists in air, water, soils or other media atlevels that present an unacceptable risk to the environment.
a. Human Health Risks
The DERR has not developed media specific action levels forchemical constituents. Therefore, a human health risk assessmentmust be performed in order to evaluate current and potentialfuture health effects posed by site specific contamination. Abaseline risk assessment must be performed to estimate thecurrent or the potential future risks presented under a no actionscenario. A no action scenario assumes that there are nocontrols, current or future, at the site (i.e., fences, deedrestrictions). The DERR has adopted U.S. EPA's Interim Final"Risk Assessment Guidance for Superfund. Volume I. Human HealthEvaluation Manual (Part A)" for conducting risk assessments atDERR unregulated sites. Appendix A lists additional guidance
necessary for conducting risk assessments.
The DERK relies on the following information sources for thetoxicological data necessary for conducting risk assessments:
* IRIS - U.S. EPA!s Integrated Risk Information System, IRIS isan electronic database containing the latest descriptive,quantitative and U.S.EPA regulatory information on chemicalconstituents. Chemical files maintained in IRIS containinformation relating to noncarcinogenic and carcinogenic healtheffects.
IRIS is accessible by means of Dialcom Inc.'s Electronic KailTelecommunication System and the Computer Information System(CIS). For information on IRIS contact U.S.EPA's Office ofHealth and Environmental Assessment, (202) 382-4317, or theOffice of Solid Waste, Characterization and Assessment Branch,Washington, D.C. (202) 382-4761.
Information in IRIS supersedes all other sources. If informationis not available in IRIS the sources given below should beconsulted.
* HEAST - Health Effects Assessment Summary Tables are citedreferences which are updated quarterly by U.S. EPA. HEAST is atabular presentation of toxicity information and values forchemicals for which Health Effects Assessments (HEA's), Healthand Environmental Effects Documents (HEED's), Health andEnvironmental Effects Profiles (HEEP's), Health AssessmentDocuments (HAD's), or Ambient Air Quality Criteria Documents(AAQCD's) have been prepared. For information pertaining toHEAST, contact the National Technical Information Service (NTIS),(703) 487-4780.
* EPA Criteria Documents - These documents include drinking watercriteria documents, drinking water Health Advisory summaries,ambient water quality criteria documents, and air qualitycriteria documents, and contain general toxicity information thatcan be used if information for a chemical is not availablethrough IRIS or the HEAST references. Criteria documents areavailable through NTIS at the number given above.
i. Carcinogens
In assessing the carcinogenic potential of compounds, U.S. EPA'sHuman Health Assessment Group classifies these compoundsaccording to the weight-of-evidence from availableepidemiological and animal studies. Compounds are classified aspossible, probable or known human carcinogens based on a systemdeveloped by U.S.EPA from the approach taken by the InternationslAgency for Research on Cancer (IAKC).
Group A Human Carcinogen
Group Bl or B2 Probable Human Carcinogen
Bl - limited human dataare available
B2 - sufficient evidence in animals andinadequate or no evidence in humans
Group C Possible Human Carcinogen
Group D Not classified as tohuman carcinogenicity
Group £ Evidence ofnoncarcinogenicity forhumans
Calculated risks are assumed to be additive between compounds andcumulative across routes of exposure. This approach assumes thatthere is no synergistic or antagonistic chemical interactions andall chemicals have the same endpoint, cancer.
In assessing the additive and cumulative effects of carcinogens,the DERH requires that Group A and Group B carcinogens beevaluated. Group C carcinogens shall be evaluated on a case-by-case basis. The DERH shall consider the appropriate site-specific environmental data and toxicological informationavailable in determining whether or not Group C carcinogens willbe included in evaluating risks.
ii. Noncarcinogens
In assessing the noncarcinogenic risk of systemic contaminants, areference dose, or RfD, is used. Additionally, one-day or ten-day Health Advisories (HA's) may be used to evaluate short termoral exposures. These values are based on subchronic and/orchronic animal studies and human epidemiological data, whereavailable. To assess the overall potential for noncarcinogeniceffects posed by multiple chemicals, a Hazard Index (HI)approach must be used. This approach assumes that simultaneoussubthreshold exposures to several chemicals could result in anadverse health effect. It also assumes that the magnitude of theadverse effect will be proportional to the sum of the ratios ofthe subthreshold exposures to acceptable exposures.
It is important to calculate the Hazard Index separately forchronic, subchronic and shorter term exposure periods. Onceagain, the Hazard Index approach assumes additivity betweencompounds and cumulative effects across exposures routes. Theassumption of dose additivity is most appropriately applied tocompounds that induce the same effect by the same mechanism ofaction. Therefore, in applying the Hazard Index approach thetotal Hazard Index should be calculated. If the total HazardIndex exceeds unity it is appropriate to segregate the compoundsby effect and by mechanism of action and to derive separatehazard indices for each group.
Example;
Hazard Index - fcl + £2 + . .. . EiRfDi RfD2 RfDi
Where Ei - Exposure level (or intake) for the ith toxicantRfDi - Reference dose for the ith toxicant
b. Environmental Risks
Where contamination is contributing to soil, sediment, air and/orwater pollution, the DERR requires an evaluation of theenvironmental impact of such contamination. Environmentalimpacts are evaluated by assessing adverse effects on the floraand fauna existing within or threatened by contaminated media(air, water, soils or sediments). Specific data must beavailable in order to establish the nature and magnitude of thecontamination and its effect(s) on the environment. The siteinvestigation must include the following in order to evaluatethese effects:
* Appropriate chemical and physical data describing thenature of contaminants and the contaminated media;
* Ecological assessment (e.g., habitat, variability inpopulation and diversity of species, food chaineffects) through toxicity tests, biomarker analysis andfield surveys;
* Toxicological data in regard to flora and fauna;
* Evaluation of especially sensitive habitats andcritical habitats of species protected under theEndangered Species Act and habitats of Ohio EndangeredSpecies as determined by the Ohio Department of NaturalResources, Division of Wildlife.
The DERH requires that an ecological assessment be evaluated foreach site. The Division recognizes that there will be instanceswhere ecological impact may be small or insignificant due to thenature of site contamination or the location of the site. Insuch instances it must be demonstrated, either qualitatively orquantitatively, that significant environmental impacts have notor will not occur and a full ecological assessment is notnecessary.
The DERR adheres to the procedures presented in U.S.EPA's"Ecological Assessment of Hazardous Waste Sites: A Field andLaboratory Reference" EPA/600/3-89/013, and "Risk AssessmentGuidance for Superfund, Volume II. Environmental EvaluationManual" EPA/540/1-69/001 when conducting ecological assessments.Additional references for conducting ecological assessments arelisted in Appendix A.
III. EVALUATION OF ALTERNATIVES
Once site contamination has been characterized and the threatposed by the contamination has been determined, remedialalternatives must be developed and evaluated if the contaminationis found to present an unacceptable risk to human health or theenvironment or if any promulgated standard or criteria has beenviolated. The development and evaluation of alternatives can beintegrated with site characterization activities. The DERRrequires that remedial alternatives be developed that areprotective of human health and the environment. Remedial actionalternatives shall be evaluated using the the following criteria:
1) Overall protection of human health and the environment2) Compliance with applicable or relevant and appropriatestandards and/or criteria3) Long term effectiveness and permanence4) Reduction of toxicity, mobility, or volume through treatment5} Short term effectiveness6) Implementability7) Cost8) Community Acceptance
Alternatives should establish remediation goals that meet thecriteria outlined in the following sections.
A. Promulgated Standards
Remediation goals must consider any cleanup standards, standardsof control, and other criteria or limitations promulgated underfederal or state environmental or facility siting laws thatspecifically address the circumstances at the site.There may be other requirements, criteria, or limitationspromulgated under federal or state environmental or facilitysiting laws that, although not directly applicable to the site,are sufficiently similar to be suitable to the situation.
There are several different types of requirements promulgatedunder State or Federal law which may apply:
* Ambient or chemical specific requirements - Thesevalues are usually health or risk based numericalstandards or criteria and establish the acceptableamount or concentration of a chemical that may be foundin, or discharged to, an environmental media.
* Performance, design, or other action-specificrequirements - Technology or activity basedrequirements or limitations.
* Location specific requirements - Restrictions placedon the concentration of a substance or the conduct ofactivities because of the location of the site orrelease.
The DERR has developed separate policy for the identification andapproval of promulgated standards to be applied at unregulatedhazardous waste sites. This policy was specifically developedfor the identification of state applicable or relevant andappropriate requirements (ARARs) at National Priority List (NPL)sites. However, the policy procedures are applicable to non-NPLsites as well. This policy, entitled "ARARs" is availablethrough the DERR.
Maximum contaminant level goals (MCLGs), established under theSafe Drinking Water Act, shall be attained by remedial actionsfor ground or surface waters that are current or potentialsources of drinking water, where the MCLGs are relevant andappropriate under the site specific circumstances. Where an MCLGfor a contaminant has been set at a level of zero, the maximumcontaminant level (MCL) promulgated for that contaminant underthe Safe Drinking Water Act shall be attained where the MCL isrelevant and appropriate under site specific curcumstances.
In cases involving multiple contaminants or pathways, chemicalspecific KCLs or MCLGs must meet a cumulative carcinogenic riskof 10-4 to 10-6 and Hazard Index less than one in order to beacceptable as remediation goals. This demonstration must beperformed through the baseline risk assessment conducted for thesite.
B. Acceptable Risk Criteria
If state or federal promulgated standards do not exist for acontaminant or contaminants in a specific environmental media, orif standards are not sufficiently protective due to multiplecontaminants or pathways, remediation goals shall be establishedto meet risk based criteria. The following criteria are deemedprotective of public health and the environment.
a. carcinogens - For Group A and Group B carcinogens acceptableexposure levels are generally concentration levels that representa cumulative excess upper-bound lifetime cancer risk to anindividual between 10-4 and 10-6. The 10-6 risk level shall beused as a point of departure for DERR remediation goals.
b. noncarcinogens - For noncarcinogens, acceptable exposurelevels are generally those levels which represent concentrationsto which the human population, including sensitive subgroups, maybe exposed without adverse effects during a lifetime or part of alifetime. Cumulative exposures which present a Hazard Index ofless than one are considered acceptable for site remediationgoals.
c. environment - Acceptable levels are generally those levelswhich represent concentrations to which the ecologicalenvironment, including sensitive species, habitats and criticalhabitats, may be exposed without adverse effects.
In order for a site to be considered unrestricted for future use,
cumulative carcinogenic risks oust meet the 10-6 goal andcumulative noncarcinogenic risks must have a Hazard Index belowone. Unrestricted use assumes that no future monitoring,controls or restrictions will be required at the site.Therefore, it must be demonstrated that any remainingcontamination left on site will not pose an unacceptable risk tohuman health or the environment.
The carcinogenic risk range allows for some flexibility inestablishing cleanup goals that are protective of human healthand the environment through applying a more restrictive usescenario. These scenarios may include long term monitoring andsite controls which will place restrictions on any wastes left onsite and limit future use of the site. Such alternatives »ay beconsidered based on site specific considerations and if anunrestricted future use scenario is technologically infeasible orcost prohibitive.
Remedies evaluated should use treatment to address contaminationwherever practicable. The use of innovative technologies isencouraged when such technology provides for comparable orsuperior treatment performance or implementability. The DERRwill require that environmental media be returned to usable andbeneficial uses wherever practicable.
IV. SUMMARY
The DERR, Remedial Response Program is responsible for thediscovery, listing, prioritization, investigation and potentialremediation of over 1300 unregulated sites in Ohio. These sitesmay be vastly different in regard to the nature and extent ofcontamination. Every site presents a unique set of circumstancesthat must be addressed individually. The DERR has developed thispolicy to provide a consistent approach to site investigation andremediation under the Remedial Response Program. This policy isintended to provide general guidance to DERR staff and theregulated community. This policy is not intended to be the solesource of guidance, it is intended to be used in conjunction withother DERR approved guidance and policy.
The essential steps in any site evaluation include:
1) Site Investigation - the determination of the nature andextent of contamination through on-site and backgroundcharacterization.
2) Risk Assessment - the determination of current or potentialrisk to human health and the environment by evaluating risksposed by on-site contamination.
3) Evaluation of Alternatives - the determination of a sitespecific remedial action by developing alternatives which areevaluated against applicable or relevant and appropriatepromulgated standards, human health and environmental criteria.
10
The DERR will adhere to the above steps when conductinginvestigations and clean-ups at unregulated sites or inconducting oversight at sites where action is being taken byanother party under an enforcement agreement.
Due to the large universe of unregulated cites listed by the DERRfor potential action, the Remedial Response Program will not beable to address every site as quickly as interested parties naywish. The Remedial Response Program prioritizes sites in orderthat the sites presenting the greatest risks be addressed first.Sites are addressed as resources and funding allow. The DERRwill only conduct oversight at sites where the Agency has aformal agreement in place. Parties may wish to proceed with asite assessment and remediation on their own, without any DERRoversight. The How clean is Clean policy, in conjunction withother DERR guidance and criteria (Appendix A) should be used asguidance in conducting these investigations.
11
ir c •>!*•>!*! s k .•
?x:- *u: "ct! i . ' ; ;^ "?:•:•.:: :? !•?{:: •• T: : fc-:;::-. :E :: & : i t c? K C ; : ? ? ::-«
*HO FU*WR
|y PEFR
wu'i:ic« orMXiCiTc:s?;s o f c »ennui.i N ( - t >M (7E&!{ V f i l V f i T IC K I U K . f l
.f s:;s.
APPENDIX A
REFERENCES
1) ARAR's, Ohio EPA, Division of Emergency and RemedialResponse, July, 1991.
SOURC: (t) for itaole sizes i 50: Utbennan, Gerald F. 1958. 'Ttblw forSUtmical Tolerance L1«1ts.' Industrial Quality Controlle I1m * 50: ' vtlu"
B-9
The values -1.163, -.495. 0. .495. and 1.163 are called expeaed normalscores lor a sample ol size five trom a standard normal distribution. The val-ues .488, .495, .500, .505, and .512 are the expeaed normal scores for asample of sire five from a normal distribution with M • .500 and a * .010.In general, whatever the values of M and <r, the smallest expected normalscore in a sample of size Eve is M + (-I.l63)a, the second smallest expeaednormal score is p •*• (-.495)<7, etc. Once the expeaed normal scores forM " 0, a - 1 are available, those for any other values of M and a are easilycalculated.
For example, Table II shows that when n « 25, the third smallest expeaednormal score when M - 0, a « 1 is -1.263. If observations are taken ingroups of 25, the long-run average of the third smallest one will be roughly-1.263. If height is normally distnbuted with M • 69 in., tr « 3 in., andheight values are obtained in groups of 25, the long-run average value of thethird smallest height value would be 69 * 1-1.263M3) - 65.211.
A Plot ior Chcclune The expeaed normal scores for a sample of size n - 5 when M * 50,Normals " a - 10 are 50 •*• (-1.163K10) - 38.37, 45.05, 50, 54.95, and 61.63. Sup-
pose we obtain a sample of five x values from a particular population undersrudy. If the population distribution is actually normal with M « 50, cr » 10,
_ __ _then the five observed scores should be reasonably close to the expeaed nor-mal scores. Consider the "pain (smallest expeaed normaTscore, smallest"observed value), (second smallest expeaed normal score, second smallest ob-served value), . . . , (largest expeaed normal score, largest observed value). Ifevery observed value coincides exactly with the corresponding expeaed nor-mal score, the resulting pair* are (38.37. 38.37), (45.05. 45.05), (50, 50),154.95. 54.951, and (61.63, 61.63). Figure 14(a) shows a plot of these pointswiih the horizontal axis identified with expeaed normal score and the verti-cal axis identified with observed value. The five points fa l l exaaly on a 45*line passing through the point (0, 0).
1 50H
I
ROUU14
Exptcttd
(b)
0 1Expect*) /
(c)
NORMAL PROBABILITY PLOTSHi Whtn Obttrvtd Vilut* • Exptcttd Scornibl Whtn Obitrvtfl Valuti Art Clou to Exptcttd Scorn,Ci Whtn ExptcitO Standard Normal Scortt Art Uttd
11 O4CXMG K* MOAMAUTY 211
In practice, even when the underlvmg distribution is identical 10 the normaldistribution for which the expeaed normal scores are computed (heren * iO . t r - 10), observed values usually do not coincide exactly with theexpeaed normal scores. The observed scores (ordered) might be 39.2, 43.5,49.1. 55.4, and 55.9, yielding the texpeaed, observed) pairs (38.37, 39.2),<45.Q5. 43.5), (50. 49.1). 154.95. 55.4). and (61.63, 59.9). Figure H(b) dis-plays a plot of these pairs. The points fall reasonably close to the 45* line.
Suppose that instead of plotting observed versus expeaed normal scoreswhen M » 50, Q • 10, we plot observed versus expected nonnal scoreswhen M " 0, y • 1 (the standard normal expeaed scores). For our example,this involves plotting (-1.163, 39.2), (-.495, 43.5), (0, 49.1), (.495. 55.4),and i1.163, 59.9). Figure 14(c) shows the resulting plot. It has exactly thesame general pattern as the plot of Figure 14(b); the points fall close to astraight line, but it is no longer the 45* line of the two earlier plots. This sug-gests that plotting observed values versus expeaed scores when M " 0,a = I will yield a straight line pan em if the underlying distribution is nor-mal irrespective of the aoual values of M and cr, and this is indeed the case.
To construa a plot for checking normality, order the n sample observa-DODS from smallest to largest, obtain the expeaed normal scorn whenM " 0, g * I from Table 11 (or some other more complete source), and
"TornTIhe n (expeaed score, observed valueT pain. If "the distribution"from which the sample was obtained is normal (at least approximately),a plot of these pain should show a reasonably strong linear pattern. Ifthe underlying distribution is disonaly nonnormal, the plot should showa rather pronounced departure from lineanry. The plot is usually called anormal probability plot.
Figure 15 contains four such plots for samples of n « 25 observationseach. The first plot shows the expeaed straight-line pattern consistent with anormal distribution. The second plot is the son of picture that rypicallv re-sults from sampling a distribution that is symmetric but has heavier tails thanthe nonnal curve. The middle pan of the plot is reasonably linear, but on theleft end the points fall below a straight line through the middle pan and onthe nght end the points fall above such a straight line. This difference occurssince, because of the heavy tails, observations at the upper end tend to belarger than what it expeaed from a nonnal distribution. Correspondingpoints at the upper end of the plot then tend to be higher than those resultingfrom a normal distribution. Figure 15(c) illustrates what would occur if theunderlying distribution had lighter tails than a nonnal distribution. The na-ture of the departure from lineanry at extreme ends of the plot is the mirrorimage of what happens in the heavy-tailed case. The plot in Figure 15(d) istypical of what results from sampling a distribution that is quite skewed—arather strong curved pattern in the plot.
In (he best of all possible worlds, the points in a nonnal probability plotfall exaaly on a straight tine, but in practice, sampling variability precludessuch an ideal picture. How far can the pattern in the plot deviate from linear-ity before the assumption of normality should be judged implausible? This is
212
-1 0Expcci«d
dl
60
SO
40
X
20- 2 - 1 0 1
Expccttd /
(b)
45
40
-60-> 50"
30-20-
30
25
AOUAC
-2 -2 -1
(c) (d)
NORMAL PROBABILITY PLOTS(•i Obiervitions from • Norm*I Diftnbutionib) Obtcrvaiioni from • HMvy-T»iltd Divtribulionic) Ob*trvfciiont from • Light-T»tl«o OittnbutionId) Obitrvtiiont from • Pontivtly Sk«w*d Diitnbution
not an easy question to answer. To get a reeling for what die plot might looklike when the distribution is normal, for each of a number of different valuesof the sample size rt, we could generate a number of samples of n normallydistributed observations and study the resulting plots. This obviously requiresa large investment of space (for us) and time (for you), both of which art inshort supply. The book Fining Equations to Data by Cuthbert Daniel andFred Wood (New York: John Wiley, 1980) presents a number of cucfa plotson pages 33-43. Tliese plots suggest that with small ample sizes (e.*.,n < 20), then can be so much sampling variability that substantial depar-tures from linearity can result even when the distribution is normal. So be
U O«CK»0 219
EXAMPLE 12
careful about deciding against the plausibil i ty ot a normal distnbution basedon a normal probability plot when n is small.
As previously mentioned, the computer is very good at constructing normalprobability plots. In particular, all the most frequently used packages ottuasncal computer programs have a command that will produce such a plot.The details of the plot vary somewhat from package to package—for exam-ple. MlNTTAB and BMDf use (different! internally calculated approximations tothe expected normal scores rather than tabulated values—but the key idea isalways TO look for linearity.Example 29 of the previous section referred to an article that containedn • 46 observations on the amount of oxides of nitrogen (NO.) emitted by aparticular type of automobile. Figure 16 shows a normal probability plot ofthe data produced by MINITAB. Where two points in the plot fell so close toone another that separate asterisks would not show, the charaaer 2 was usedinstead, and three close points were replaced by the charaaer 3. The plot has'a reasonably well defined straight-line charaaer, though there is a bit of wob-blinns in the two tails. With only a moderately large sample site, such wob-bling is quite common. We conclude that an assumption of normality for thedistribution of NO. emissions is quite plausible.
3 .30*
2.10*
Amnt0*
NO.
221 .40*
0.70*
• 2222
noimc it
0.00*
• 2.5 • 1.9 .0.5 0.9 1.9 Z.9
MINITAB GENERATED NORMAL PROBABILITY PLOT FOR NO, EMISSION DATA
EXAMPU13 Example 18 of Chapter 2 presented data on cell interdivision times (IDT'S). Asample histogram of the original data was quite skewed, but transforming bylogarithms yielded a reasonably bell-shaped histogram. Figure 17 displays