> NPMc bjbj W\W\KY%NNt:::\^4:*!^"m o o o o o o $"z%p Q !!!jm !m !!!]ovW!Y 0*!!%]%!%!8! ?*!%N>: Glossary of Terms
accuracy - The proximity of a quantified measurement result to the true value of the property being measured. Every property being measured (the measurand) has a theoretically true value. The extent to which a measurement result approximates the true value is an assessment of accuracy.
ANOVA - analysis of variance (ANOVA) is a statistical method designed to compare inferentially the differences between groups or levels of some factor of interest. The analysis is done by comparing the variance between levels to the variance within levels for the factor. There are both one-way and two-way ANOVA models. Interaction between factors in a two-way model can also be evaluated. Various assumptions (e.g., equality of variance between levels) accompany the ANOVA model.
Bayesian statistics - named for Thomas Bayes (1702-1761) the English cleric and mathematician. An approach to inferential statistics where the parameter is considered a random variable and the observed experimental data are considered fixed. The observed experimental data (called the likelihood) are combined with prior information regarding the parameter (the prior distribution) and we obtain a posterior estimate (the posterior distribution) for the parameter. The confidence in the posterior estimate of the parameter is measured with a credible interval (the analog to the classical confidence interval). Classical statistics is the counter interpretation of inferential statistics.
bias - The quantitative assessment of accuracy. The magnitude of deviation between the measured result and the true value of the property being measured. Bias is caused by fixed or deterministic properties (e.g., calibration, etc.) that are constant with every measurement. Bias is frequently stated in terms of percent although the units of the measured property can also be used. Systematic error is another term commonly used to describe bias.
Classical statistics - an approach to inferential statistics where the experimental data are considered random and the parameter is considered fixed. Once the experiment is performed a confidence interval is constructed to evaluate the estimate of the parameter. Classical statistics are by far more widely employed in science than are Bayesian methods.
Coefficient of variation - Also known as the relative standard deviation, this is a measure of relative variation. The percent CV is determined by dividing the standard deviation by the mean and multiplying by 100.
combined standard uncertainty An estimated standard deviation determined from the square root of the sum of the several variance components that contribute to the measurement uncertainty.
confidence interval - An interval that is symmetric about some sample statistic (e.g., the sample mean). Since the sample mean is a random variable so are the confidence interval limits. The population parameter (e.g., the mean or ) is some unknown but constant value. A confidence interval will bracket the population parameter with some probability (in the long term frequentist sense). The limits of the confidence interval are functions of the desired confidence, the variability, and the sample size. Confidence intervals are very useful for quantifying and providing the direction of measurement uncertainty.
control chart - A sequential plot of measurement results over time allowing a visual evaluation of the measurement system. Generally, control charts will plot summarized measurement results (e.g.,
means, ranges, standard deviations, etc.) rather than individual
results. Pre-specified control limits determined as functions of the measurement variability are included in the plot to visually identify when the process is in or out of statistical control.
correlation coefficient - a measure of the linear association
between two variables. The correlation coefficient varies from
-1.0, having a negative slope, to +1.0, having a positive slope.
Values of -1.0 or +1.0 have perfect linear correlation and all of
the data lie on the line. The correlation coefficient, commonly
associated with simple linear regression, does not measure
agreement between the two methods but only association.
covariance - a measure of how two variables jointly disperse or
co-vary within the same set of observations.
degrees of freedom - the number of observations that are allowed
to freely vary or take on any value. This is determined by the
constraints put on the use of the data. The mean, for example,
has n degrees of freedom. All of the data (n observations) are
allowed to be any value along the real number line. If we apply
the constraint that the sum of n observations must add to ten then
we have n-1 degrees of freedom. n-1 observations can be any
values, but once they are determined then the last one (the nth
observation) is fixed so the values all sum to ten. The variance
has n-1 degrees of freedom because the sum of all of the
differences between the individual observations and the sample
mean are constrained to add to zero. n-1 observations are allowed
to take on any values while the last value is then constrained to
ensure the sum of the differences add to zero.
dry-gas standard - Gas mixtures containing ethanol and some inert carrier gas (e.g. nitrogen) mixed to known concentrations and used to calibrate and test breath alcohol instruments. These gas mixtures do not contain water and are thereby less representative of actual human breath.
EPROM - An acronym for Erasable Programmable Read Only Memory. This device (typically in the form of a chip) is used in many modern breath alcohol instruments to store the software or program used by the microprocessor in the operation and control of the instrument. The memory of the EPROM is removed by exposure to ultraviolet light, allowing it to be reprogrammed.
error - Those deviations possessed by all measurement results as they approximate some true property of interest. Measurement error can arise from many systematic and random causes. In the context of measurement, error and uncertainty have essentially the same meaning. The important consideration with error is that we identify it, quantify it and then minimize it for a particular measurement context.
error propagation - Measurement results are the product of a process involving many steps (e.g., signal processing, computations, transformations, etc.). Each one of these steps will introduce its own unique elements of uncertainty or error. These individual contributions of error are passed along (propagated) in each step so the final measurement result possess the accumulated error of many contributing elements.
expanded uncertainty An interval resulting from multiple (usually k=2) of the combined uncertainty and within which the measurement value is expected to lie with some stated level of probability.
experimental unit The smallest sub-division of the experimental material that is capable of receiving one of the treatments. Usually, the experimental unit is an individual person, an animal, a bottle of solution, etc.
external standards - These generally consist of either wet-bath simulator devices containing an ethanol and water solution heated to a known and constant temperature or a compressed dry-gas (ethanol and inert gas mixture) cylinder. These possess known and independently determined concentrations of ethanol which are external to the instrument and serve as a standard for determining accuracy and/or precision of the instrument.
fit-for-purpose - Every measurement has the purpose of providing numerical information to facilitate an informed decision within some pre-specified context. Some measurements have trivial applications and minimal consequences. Others have very critical applications with serious consequences. The measurement algorithm must be specifically designed in view of the purpose and application of the results. The complexity of the measurement algorithm and need for controls must be appropriate for the measurement's applied purpose.
fixed effects ANOVA model - An ANOVA model where the factor levels consist of specific chosen levels of interest. For example, the levels might be gender, specific instruments, specific hospitals, etc. The levels are not considered a random sample of all possible levels for the factor. The fixed effects model is also known as a Model I ANOVA.
flash EPROM - A type of non-volatile memory storage device used in some modern breath alcohol test instruments to store the software. Rather than relying on ultraviolet light to clear its memory, a flash EPROM device can have its memory cleared electrically either in part or in whole. This would allow the software in an instrument can be altered from a host computer via modem.
forensic integrity - Scientific evidence introduced in criminal trials has significant implications for the defendant. Since forensic scientists know that their analytical results will be closely scrutinized prior to or during the trial they generally employ highly credible methodology and protocols. The implications of forensic evidence generally requires that the highest possible standards for quality control be employed so the evidence can be received by the court with appropriate confidence.
Gaussian - a name commonly given to the normal distribution in honor of Karl Gauss, the German mathematician who developed much of the theory underlying the distribution.
hardware - This classifies the actual physical components within an analytical instrument including: sample chamber, detectors, circuit boards, printers, power supply boards, valves, etc.
hypothesis - an assumption made regarding a parameter (i.e., a mean, variance, slope of a regression line, etc.) of interest so an experiment can be designed and a statistical test performed. One might hypothesize that the within-subject mean blood alcohol concentration is equal to the within-subject mean breath alcohol concentration. This would be known as the null hypothesis and stated as: EMBED Equation.3 . There is always an alternate hypothesis as well which could be stated, in this example, as: EMBED Equation.3 . An experiment would then be designed and performed to determine which hypothesis the evidence supported. Hypotheses are never proved to be true. Instead, one simply accumulates evidence for the probable truth of the hypothesis. On the other hand, one can disprove a hypothesis by demonstrating experimentally that the hypothesis cannot be true.
interfering substance - Volatile organic compounds potentially present in human breath at concentrations exceeding the LOD of the total analytical method that could contribute to the measured concentration of ethanol. These may be of endogenous or exogenous origin. Most manufacturers incorporate features that will detect significant concentrations of acetone to avoid its contribution to the analysis of ethanol.
internal standards - Hardware devices typically of an optical nature that are introduced into the path of infrared instruments to ensure the proper response and signal processing. They help to ensure correct calibration and proper analytical operation of the instrument. Internal standards may be quartz filters or metallic devices introduced into the infrared path.
kurtosis - a measure of the peakedness of a distribution. The normal distribution has a kurtosis of zero. A very peaked distribution would have a higher positive kurtosis value. A flat distribution would have a negative value for the kurtosis.
limit of detection (LOD) - The lowest concentration at which a particular measurement system (including instruments, protocols and sample source) can distinguish between a normal background blank result (noise) and the analyte of interest at some specified level of probability. The LOD is generally equal to 3S0 where S0 is the standard deviation of the distribution of blank or background noise responses of the instrument.
limit of quantitation (LOQ) - The concentration at which the measurement system can provide results capable of being accompanied by confident estimates of systematic error or bias. The LOQ is generally equal to 10S0.
matrix all of the components contained within the test sample excluding the analyte of interest.
measurement algorithm - The detailed step-by-step process by which some measurand (measured property) of interest is collected and processed through to where the results are interpretable and informative for some decision maker. The measurement algorithm will include all of the instrumentation, personnel and protocol necessary for the complete measurement process to occur. Its complexity depends on the purpose of the measurement.
mouth alcohol - Vapor ethanol originating in the oropharyngeal region due to recently consumed alcoholic beverage or regurgitation of alcohol contents from the stomach. Since breath alcohol measurement is designed to quantify the ethanol originating in the respiratory system, any mouth alcohol that exceeds the end-expiratory sample would bias the result high. Fifteen minutes is more than adequate to ensure that mouth alcohol will not bias the end-expiratory breath sample.
noise - A term generally used to describe measurement interference of any type. Noise is any analytical signal produced from some source other than that of the analyte or property of interest. Noise might be from electronic sources, other volatile organic compounds, heat, vibrations, radio frequency interference, etc. Measurement systems should be designed to have large signal-to-noise ratios.
normal distribution - probably the most commonly applied and understood parametric probability distribution. The normal distribution is characterized by the symmetric bell shaped curve. Two parameters fully characterize the normal distribution, the mean and the variance 2. Many biological and natural phenomenon are distributed normally. Commonly used statistical tables are based on the standard normal which has =0 and 2=1.
outlier - a discrepant value in a group of measurements that does not conform to the rest. An outlier is usually identified by its large deviation from the rest of the group of measurements. There are statistical tests available for inferentially identifying an outlier. An outlier value may be due to a significant error in measurement or simply due to measuring an individual that is not representative of the population from which the other measurements have arisen. Outliers should not be automatically eliminated but should be studied further to determine their cause.
parameter - the true value for the properties that fully define a specific distribution. For example, the normal distribution has two parameters, the mean and variance. This is noted by: EMBED Equation.3 . In most experiments, we are interested in identifying the values of these parameters. Since we never measure the entire population we measure a sample and obtain estimates of these parameters, the sample mean EMBED Equation.3 and the sample variance EMBED Equation.3 . In classical statistics, these parameters are considered fixed but unknown.
precision - The degree to which replicate measurement results agree amongst themselves. These replicate measurements must be of the same property, from the same sample source, using the same equipment, same operators, same protocol, same general time, etc. Several statistics are used to quantify precision such as: standard deviation, variance, range, coefficient of variation, etc. Precision is a very important property of measurement that must be shown to be fit-for-purpose.
probability the quantitative assessment of the confidence that we have in the truth of a particular proposition or realization of an event.
quality assurance - The process by which measurement results are shown to possess appropriate and sufficient properties that are fit-for-purpose. These properties include: accuracy, precision, analytical linearity, specificity, protocol compliance, etc. Quality assurance is typically demonstrated by showing compliance with all pre-specified analytical protocol and standards.
quality control - The method by which a measurement process is monitored over time to ensure the system remains in statistical control. Within-run tests on subjects can be evaluated to ensure compliance with all protocols and standards. Control charts are effective for visual assessment of quality control and allow for quick intervention where necessary.
random effects ANOVA model - where the levels for the factor of interest in the ANOVA model are considered a random representation of all possible levels for that factor. For example, where the levels consist of four simulator solution batches selected randomly from 100 batches made by a particular lab during one year. The random effects model is also known as the Model II ANOVA.
repeatability - Replicate measurements performed by one analyst on one piece of equipment following the same protocol and using the same sample source within a short period of time.
reproducibility - Replicate measurements that may be performed by one piece of equipment over a longer period of time and by different analysts and employing similar but different sample sources. This may also describe replicate measurements using the same material and same method but in different laboratories.
robust - A property of a measurement system where the occurrence of significant error or other unpredictable aberrations will either be detected or will not interfere with the confidence in the results obtained. The tendency for a measurement system to remain in statistical control even when outside influences (noise, etc.) might interfere with its reliable operation.
simulator - A device containing approximately 500ml of an ethanol/water solution heated to a known and constant temperature and designed to provide a known vapor concentration of ethanol for calibration and testing of instruments. Based upon Henry's Law, the vapor ethanol concentration will be known when the solution has a known concentration and constant temperature. Simulators will have associated tubing, switches, thermometers, heater and temperature regulating features.
skewness - a measure of the symmetry of a distribution. The normal distribution has a skewness value of zero. A distribution with the tail extending to the right has a positive skewness value. A distribution with the tail extending to the left has a negative skewness value.
software - This includes the set of instructions that the computer (microprocessor) portion of an analytical instrument employs as it controls various features of the instrument. The microprocessor will only perform functions as directed or instructed by the software code. The software is encoded in the form of binary (1/0 or high/low) voltages. A specified sequence of these binary states provides an instruction for the microprocessor. Computerized breath alcohol instruments rely on software written on other computers and then stored in ROM (read only memory) devices within the instrument.
specificity - Many analytical systems are designed to identify and quantify a specific analyte (e.g., ethyl alcohol or ethanol) and minimize the risk of having other compounds (interferants) add to the result. The extent to which an analytical system is capable of reliably identifying and quantifying a specific analyte is an assessment of its specificity. Many analytical and protocol features have been employed to enhance the specificity of breath alcohol analysis for ethyl alcohol.
standard uncertainty When the measurement uncertainty is expressed as the standard deviation.
statistical control - Measurements performed on a system over time will be expected to vary randomly by chance causes. Every measurement system has inherent properties regarding accuracy and precision which can be determined when the system is operating within statistical control. The system can then be observed over time to determine if its variability is what would be expected by chance or if it has lost statistical control due to some causal factor. The control chart is the principle tool for monitoring a system's statistical control.
systematic error - Measurement error due to fixed or constant sources that are the same in magnitude and direction for all measurements. Bias is also used to describe systematic error. Instrument calibration is a common source of systematic error. Again, the systematic error needs to be acceptably minimized for a particular measurement context.
traceability A property of measurement results whereby it can be related to national standards (i.e., NIST) through an unbroken chain of comparisons, each having their stated uncertainty. Traceability is important to ensure accuracy and comparability of measurement results.
uncertainty - The quantitative interval within which a measured result is expected to lie. A quantitative assessment of the dispersion in the form of a range. All measurements possess uncertainty due to limitations in technology and methodology. Imprecision contributes to uncertainty. No measurement is perfect. The important thing is that the uncertainty be known and minimized so the process is fit-for-purpose. The uncertainty is quantified by the standard deviation and usually denoted, for example, by the value uy (meaning the uncertainty in the measurement y or the standard deviation of y).
variability - the extent to which replicate measurement results (e.g., data) show dispersion or disagreement among themselves. This dispersion can have a variety of causes. Var6 < 46NW3K
4ȸȬ|pdpdTdHhZCJOJQJ^Jh|h0Z\5CJOJQJ^Jh0Z\CJOJQJ^Jhz1CJOJQJ^JhQ9CJOJQJ^JhS~CJOJQJ^Jh\&CJOJQJ^Jhi^WCJOJQJ^JhCJOJQJ^JhYJh45CJOJQJ^Jh4CJOJQJ^Jh~/y5CJOJQJ\^Jh~/yCJOJQJ^J h~/y5CJOJQJ\^JaJ6 7
b-
Hu[\дШueUh}$hMx5CJOJQJ^Jh}$h}$5CJOJQJ^JhMxCJOJQJ^JharCCJOJQJ^Jh=CJOJQJ^Jh~/y5CJOJQJ\^Jh~/yCJOJQJ^Jh{0CJOJQJ^JhyhTV=5CJOJQJ^JhTV=CJOJQJ^Jh0Z\CJOJQJ^JhZCJOJQJ^Jh|CJOJQJ^J!acg!#,.VX`˳˳˳˳㧗{ochcK+CJOJQJ^Jh;SCJOJQJ^Jh(rCJOJQJ^Jh}$h[5CJOJQJ^Jh}$h}$5CJOJQJ^Jh[CJOJQJ^Jh\PCJOJQJ^Jh9CJOJQJ^JhCJOJQJ^Jh`:CJOJQJ^JhMxCJOJQJ^Jh)rhMx5CJOJQJ^J!-jYb(l-q6!7!""s$)XZ Uac')^km,.pr7!2?2B2x22333D5E5W5666˻˨ˇ{o`oT`oH8hGjh>5CJOJQJ^Jh>CJOJQJ^Jh%qCJOJQJ^Jh~/y5CJOJQJ\^Jh~/yCJOJQJ^JhcCJOJQJ^Jh CJOJQJ^J)jkhm[h"#CJEHOJQJU^J%jPO
h"#CJOJQJUV^JjhCJOJQJU^JhCJOJQJ^Jhj7CJOJQJ^JhHACJOJQJ^Jjhm[CJOJQJU^J6667778888^9y95:6:9:?:::R<_<2>7>??AABBBCEEEEEzF{FFܳܳܧ{k{jh0CJOJQJU^Jh0CJOJQJ^JhO0hAN5CJOJQJ^JhO0hO05CJOJQJ^JhANCJOJQJ^Jh~/yCJH*OJQJ^Jh~/yCJH*OJQJ^Jh~/y5CJOJQJ\^Jh~/yCJOJQJ^JhGjCJOJQJ^Jh>CJOJQJ^J%\9]9^98:9:::Q<R<1>2>??BBEEHHHIIuJvJKK[M\MNNFFFFfGgGzG{G|G}GGGGGGGGGGGHHǻwdOwCw7hANCJOJQJ^JhqACJOJQJ^J)j\ hN00hqACJEHOJQJU^J%jO
hqACJOJQJUV^JhuCJOJQJ^J)jLhN00hqACJEHOJQJU^J%jO
hqACJOJQJUV^JjhN00CJOJQJU^JhN00CJOJQJ^Jjh0CJOJQJU^J)jh0hN00CJEHOJQJU^J%j5O
hN00CJOJQJUV^JHH
HIIvJJKL\MvMNNOOPPHRQR)T*T2T@UAUIUWWYYYYY
\\l]m]y]^^^'``````s"h1h15CJH*OJQJ^Jh1h15CJOJQJ^Jh1CJOJQJ^JhShS5CJOJQJ^JhSCJOJQJ^JhdL?CJOJQJ^Jhghg5CJOJQJ^JhgCJOJQJ^Jh~/y5CJOJQJ\^Jh~/yCJOJQJ^J-NOOPPFRGRHR)T*T@UAUWWYYYYY\
\l]m]^^```ր׀``aր׀Ńƃǃʃ˃Ѓу҃ӃԃՃփ忳嫧wfw`
hSCJ!h&iCJOJQJ^JmHnHuhSCJOJQJ^JjhSCJOJQJU^JhS
hS0JjhS0JUh1ijh1iUh&iCJOJQJ^JhHrCJOJQJ^Jh%qCJOJQJ^Jh'CJOJQJ^JUh~/yCJOJQJ^Jh~/y5CJOJQJ\^J%iability can be quantified by several statistics including: standard deviation, variance, range, coefficient of variation, etc. Smaller variability (e.g., greater repeatability) is desired in measurement systems.
variance components - Measurement variability is the result of many factors that all contribute in random and frequently independent fashion. These components might include: power supply, electronic components, tubing, other instrumental features, sampling techniques, operator errors, etc. The magnitude to which each of these components contribute to the variability in the final result can be determined by summing each of their individual variances. Many times it is difficult to identify and measure individual components of variance within an instrument. Typically, the biological and instrumental components of variance are measured to determine their contributions to total variability.
RGG 6/8/2017
PAGE
PAGE 9
׀ǃȃɃʃԃՃփ׃$&`#$a$d&`#$փ׃h~/yCJOJQJ^Jh1i<00PBP/ =!"#`$`%Dp< 00PBP/ =!"#$%Dp< 00PBP/ =!"#$%Dp< 00PBP/ =!"#$%Dp< 00PBP/ =!"#$%Dp< 00PBP/ =!"#$%Dp< 00PBP/ =!"#$%DpkDd
,<Eb
c$A??3"`?2Oop|D`!Oop|
WxR=KP=K!%HGAh""#
Ɗ-BU:'(ũ`G=9'P_u*ҮȬ^iL8u&RyU9R@1UHìSnлhnB`_kϤKgx$$j({a=D5_zxD&XWbh2ubF<7SΙL?A뿚Fܦ8_e_:i|\s\a$O?Dc٪7_緟(8ҩ24ʺ{ \n
0Bp Y4]41)ZՠC~>.c4Jկ8k4{;9ZX
̀oy
qap]f/^^-{mpbWEoގze.VZ-|_KoSGbmc$7 `!/c(_⼨>c$xPNQ=3 .
!Xh!4[ $ {i>P,L2{f4/9-pDm.R&ZNU.d!m֯gP8hJ
+L_|d߽ј=&3uEu(~ݥo|߭Չ&> H<ߤ[6::O`src+Z0Z
>?^1gxl~q|?J
!"#$%&'()*+,-./0123456789:;<=>?@ABCDFGHIJKLORSTVUWXY[Zp\^_`abcdefghijklmnoRoot Entry FvQData
EWordDocumentObjectPool jvv_1338616764FjvjvOle
CompObjfObjInfo
!"#$&'()+
FMicrosoft Equation 3.0DS EquationEquation.39q8!
N,2
()
FMicrosoft Equation 3.0DS EquationEquation.39qOle
CompObjfObjInfoEquation Native 0<(8!
Y
FMicrosoft Equation 3.0DS EquationEquation.39q<H}
S2Oh+'0 _1338617518FjvjvOle
CompObjfObjInfoEquation Native 61Table]%SummaryInformation(DocumentSummaryInformation8%(
<H
T`hpxWashington State PatrolNormalRod Gullberg48Microsoft Office Word@@$]@bX
S
K՜.+,0hp
Breath Test Section-YTitle
F Microsoft Word 97-2003 Document
MSWordDocWord.Document.89qs2 0@P`p2( 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p8XV~ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@_HmH nH sH tH P`PNormal1$7$8$H$OJQJ_HaJmH sH tH DD Heading 1$@&CJOJQJ^JDA`DDefault Paragraph FontViVTable Normal :V44
la(k (No List<& <Footnote Reference4 @4Footer
!.)@.Page Number4"4Header
!PK![Content_Types].xmlN0EH-J@%ǎǢ|ș$زULTB l,3;rØJB+$G]7O٭VvnB`2ǃ,!"E3p#9GQd; H
xuv 0F[,F᚜KsO'3w#vfSVbsؠyX p5veuw 1z@ l,i!b
IjZ2|9L$Z15xl.(zm${d:\@'23ln$^-@^i?D&|#td!6lġB"&63yy@t!HjpU*yeXry3~{s:FXI
O5Y[Y!}S˪.7bd|n]671.
tn/w/+[t6}PsںsL.J;̊iN $AI)t2Lmx:(}\-i*xQCJuWl'QyI@ھ
m2DBAR4 w¢naQ`ԲɁ
W=0#xBdT/.3-F>bYL%˓KK6HhfPQ=h)GBms]_Ԡ'CZѨys
v@c])h7Jهic?FS.NP$
e&\Ӏ+I "'%QÕ@c![paAV.9Hd<ӮHVX*%A{YrAբpxSL9":3U5U
NC(p%u@;[d`4)]t#9M4W=P5*f̰lk<_X-CwT%Ժ}B% Y,]
A̠&oʰŨ;\lc`|,bUvPK!
ѐ'theme/theme/_rels/themeManager.xml.relsM
0wooӺ&݈Э5
6?$Q
,.aic21h:qm@RN;d`o7gK(M&$R(.1r'JЊT8V"AȻHu}|$b{P8g/]QAsم(#L[PK-![Content_Types].xmlPK-!֧60_rels/.relsPK-!kytheme/theme/themeManager.xmlPK-!R%theme/theme/theme1.xmlPK-!
ѐ' theme/theme/_rels/themeManager.xml.relsPK]
^/<HFTqY>|6t $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$'$16FH`փ134689:<=?C-s$\9N׀257;>B'''((<(>(z:::f;z;|;;;;qY::::: '!!8@0(
B
S ?,,MLYLTTKYMYNYPYQYSYTYVYWYoYrY#,
X_!(.8ktZ\ 'cf)4mp.0rx7<t|;G"@!E!""e%m%$&.&''%((())E+M+,,--^/c/90?000R2W2247455y7799w:z:;;<<==v>}>??\AbABBDDHFQF*H2HAIIIKKMMMM
PPmQyQRRTTpVxVKYMYNYPYQYSYTYVYWYoYrY333333333333333333333333333333333333333333333333333333333333333333333333333333336 ''((?(D+D+z::f;};;;MMoVoVDYEYIYJYKYKYMYNYNYPYQYSYTYVYWYoYrY6 ''((?(D+D+z::f;};;;MMoVoVDYEYIYJYKYKYMYNYNYPYQYSYTYVYWYoYrYHGf^p(f;SHA1u}$\&cK+N00O06W0+2j7`:P=TV==?dL?qA,gBarCYJ\PgPUUi^WcYM\0Z\gGj%q)r(rHrMx~/y|S~SAN14>"#Q9Vrz1'aZL9c0 1i9=&im[8yY{0[KYMY@