Glossary

  1. Primary Words

  2. A-B-C
  3. D-E-F
  4. G-H-I
  5. J-K-L
  6. M-N-O
  7. P-Q-R
  8. S-T-U
  9. V-W-X-Y-Z


Glossary Primary Words



 1. Break
 2. Break Set
 3. Decay
 4. Homogeneous
 5. Hook-up Replicate
 6. Lot
 7. Lot Control Standard (LCS)
 8. Lot Control Standard Set (LCS Set)
 9. Local Drift
10. Noise
11. Outlier
12. Reproducibility



Break
   A "break" is any interruption (coffee break, lunch break,
   end of day, power outage, etc.) which disrupts the
   physical flow of data during the chemical analysis
   process.

Break Set
   A "break set" is the set of chemical measurements
   collected in the interim between 2 breaks; in particular:

      1) all test cylinder readings collected up to the first break
         would be considered to have come from "break set 1";

      2) all test cylinder readings collected between the first
         and second break would be considered to have come
         from "break set 2"; and so forth.

   Over them lifetime of the 3- or 4-day chemical analysis,
   it is not uncommon to have 3 to 10 break sets.
With new instrumentation which is highly automated,
the number of breaks will be substantially reduced.

Decay
   Decay in a cylinder refers to the decrease in the
   true concentration of a cylinder over time (usually
   months).  Decay is an unavoidable fact-of-life for the
   "unstable" gases such as SO2, NOX, H2S, etc.; decay is
   less of a problem for most other gases

   Decay is sometimes referred to as global drift, but such
   designation is not recommended because "drift" is here
   restricted to changes (increases/ decreases) in the
   instrument response (as opposed to the true
   concentration) due to non-cylinder factors such as
   instrument, port, environemnt, day, etc.

Homogeneous
   The lot of cylinders is said to be "homogeneous" if the
   concentrations across all cylinders in the lot are not
   significantly different from both a chemical and a
   statistical point of view.  When a lot is
   homogenous, then the lot may be
   batch (= globally) certified--that is a common gas concentration
   value is determined and that common value is assigned to
   all cylinders in the lot, with a common value for the
   uncertainty on the true concentration.

   If the lot is determined to be inhomogenous (= heterogeneous)
   then either:

      1. the cylinders are individually certified with
         individual uncertainties (a time-intensive
         alternative); or

      2. the cylinders are batch (= globally) certified with
         a common uncertainty; note, however that such
         uncertainty will tend to be large because
         prediction intervals will be used rather than
         confidence intervals.

   On the other hand, if the common uncertainty is still
   small enough by chemical standards, then method 2 will be
   used; else method 1 will be used.

Hook-up Replicate
   Over the several days of the chemical analysis,
   each test cylinder is usually manaually hooked up
   to the measuring instrument at least once, and
   sometimes 2 or 3 times.  For a specific cylinder:

      1) all of the readings that were taken on the cylinder
         during the first hook-up would be considered to
         have come from "hook-up replicate 1";

      2) all of the readings collected on a specific
         cylinder during the second hook-up would be
         considered to have come from "hook-up replicate 2"; etc.

   An implicit assumption is that there is no difference in
   the measured concentrations between the hook-up replicates
   for a given gas cylinder.  This assumption should be
   statistically checked and verified.

Local Drift (= Drift)
   "Local drift" (or simply "drift") is the systematic,
   non-random (structured, deterministic) short-term (up to
   a few days) change in the measurement instrument response
   due to instrument, environment, etc.  Such deterministic
   changes are commonly linear, quadratic, sinusoidal, etc.
   Such changes are most easily detected by the use of lot
   control standards.

Lot (= Batch)
   A collection of gas cylinders is a "lot" or a "batch" if
   they were fabricated together in a contiguous block of
   time, and delivered together as a lot to be certifed.
   Typical lot sizes are 50 to 100 cylinders.

   Ideally a lot is homogenous, so that global
   certification of the lot will suffice.  Testing for
   homogeneity within a lot is a critical element in the
   statistical analysis.

Lot Control Standard (= Lot Control Standard = LCS)
   This is a representative cylinder (sometimes larger in
   size) chosen from the lot of cylinders to be certified.
   The purpose of the lot control standard is to detect,
   assess, and correct time drift.  When drift is detected,
   the test cylinder readings are corrected according the
   behavior of the readings of the LCS.

Lot Control Standard Set (LCS Set)
   For drift assessment/correction reasons, the chemical
   analysis experiment design ideally intersperces test
   cylinder readings with readings from a designated
   standard cylinder (the LCS: the lot control standard).

      1) After an initial LCS reading is taken, all test cylinder
         readings up to (but excluding) the second LCS reading
         would be considered to have come from "LCS set 1";

      2) all test cylinder readings collected between the second
         and third LCS readings would be considered to have come
         from "LCS set 2", and so forth.

   Over the lifetime of the 3-or 4-day chemical analysis,
   it is not uncommon to have a large number (e.g., 30 or
   40) LCS sets.  In the presence of drift, all test
   cylinders within a LCS set will have common corrections
   applied to them based on the values of the LCS's which
   bracket the group.

Noise
   Noise is that component of the instrument response which
   is random, unpredictable, and non-deterministic.  Changes
   in the level of the response due to changes in a factor
   settings (cylinder 1 to cylinder 2, day 1 to day 2, hook-up
   replication 1 to hook-up replication 2, etc.) would NOT
   be considered "noise".  Changes in the level of the
   response due to deterministic, structured, systematic
   time-induced effects (local drift) would also NOT be
   considered as "noise".  Noise is whatever is left over
   when all of the deterministic, predictable components
   have been accounted for and subtracted out.

   The noise of a system is most commonly observed in the
   non-constant, unpredictable sequence of instrument
   readings which result when multiple measurements are made
   under fixed experimental conditions (same cylinder, same
   time of day, etc.) The noise of a system commonly tends
   to a locally unpredictable but globally regular frequency
   pattern which is best summarized and described by
   probabilistic distribution models.  The most common
   "noise model" is normal (= Gaussian).

Outlier
   An "outlier" is an instrument response (a data point)
   which "behaves" as if drawn from a population different
   from the rest of the data.  An underlying assumption in
   the statistical analysis is that after the systematic
   factors have been accounted for, the data as a whole will
   "behave like" they came from a common distributional
   population, and the certificate value being produced will
   be a representative summary statistic of that common
   population.  Undetected outliers have the effect of
   biasing the summary statistics.

   More broadly, "outlier" may refer to either an

      1) instrument response value (as above); or a
      2) cylinder.

   Definition 2 (cylinder) is less common and to designate n
   outlying cylinder one must be explicit (e.g., "cylinder
   outlier 17", as pposed to only "outlier 17").  Definition
   1 (instrument response value) is much more common and is
   the preferred definition hen the "outlier" is otherwise
   unqualified.

   Outlier detection and eradication is an important
   component of both the chemical and the statistical
   analysis.

Reproducibility
   The chemical analysis process is said to be reproducible
   if multiple readings on a given cylinder, taken over the
   multi-day lifetime of the experiment, yield response
   values which are chemically and statistically equivalent.
   The most common method of assessing reproducibility is to
   take readings on the lot control standard (LCS) multiple
   times.  If there are no apparent differences in the
   response over these mutliple runs, then the process is
   said to be "reproducible".  If a process is not
   reproducible, then the usual cause is instrumental or
   procedural.

Click here to return to alphabetic list


Glossary A-B-C


..........A..........

ANOVA (Analysis of Variance)
   A statistical technique used to determine
   which factors significantly effect the
   response of interest (e.g., concentration).

ASCII
   A format for characters within files.  ASCII is an
   acronym for American Standard Code for Information
   Interchange.  The advantage of using ASCII format is
   portability and ease of I/O across all software
   packages.

Add in Quadrature
   A method for obtaining a summary uncertainty
   statistic which consists of square rooting the sum
   of the squared uncertainty components.  Thus, for
   example, uncertainties U1, U2, and U3 are "added in
   quadrature" to obtain a single summary statistic U
   as follows:

      U = sqrt(U1*U1 + U2*U2 + U3*U3)

Adjust Data
   To modify the contents of (typically) the original
   raw data set by removing outliers, correcting
   mis-read data, and/or by collecting more data.  The
   basis for such modification can be chemical,
   procedural, or statistical.

Analysis
   Analysis has 2 definitions: chemical analysis and
   statistical analysis:

      1) Chemical analysis consists of all
         concentration measurement operations and
         calibration operations needed to obtain raw
         concentration readings;

      2) Statistical analysis consists of all
         numeric and computational procedures (both
         quantitative and graphical) to ascertain the
         quality of the data, the validity of underlying
         assumptions, and the determination of final
         certificate values.

Analysis Issue

Area

Automate
   To identify and remove as many of the "manual" steps
   in the chemical/statistical analysis.
   Such automation requires
      1) hardware & computerization
      2) software
      3) process flowchart which
         identifies all possible "states" &
         decision points of the process along with
         detailed instructions as to what to do
         when each state occurs.

..........B..........

B0, B1, B2, ressd.
   Output statistics from the calibration regression.
   For linear regression, the model is Y = B0 + B1*X.
   For quadratic regression, the model is Y = B0 + B1*X + B2*X**2
   B0 is the intercept or the additive coefficient;
   B1 is the slope or the linear coefficient;
   B2 is the quadratic coefficient;
   ressd is the residual standard deviation
   (it is a measure of the goodness of fit of the linear or
   quadratic fit).

Batch
   A batch is the group of 50 to 100 gas cylinders
   produced and certified together as NTRM's.

   In general, batch refers to a homogeneous set.
   In regard to the gas cylinders, batch referes to
   the set of cylinders that are presumedly homogeneous
   because their production was in a contiguous block of time;
   such a batch is typically of about 50 cylinders and we speak
   of a single product batch consisting of these 50 cylinders.

   In regard to chemical analysis, "batch" may refer to
   a given replicate of all 50 cylinders.  The complete
   chemical analysis normally has a given cylinder (out
   of the 50) being processed (measured) about 3 times.
   Processing each of the 50 cylinders exactly once may
   take a few days and as such would be referred to as
   the first batch (over time).  Processing the 50
   cylinders a second time would be time batch 2.
   Processing the 50 cylinders a third time would be
   time batch 3. An implicit assumption is that there
   is no difference in the measured concentrations
   across the 3 time batches.  This assumption should
   be checked and verified.

Batch Control Standard (= Lot Control Standard)
   This is a representative cylinder chosen from the batch
   of cylinders to be certified. It will be used for intercomparison
   by the other cylinders in the batch; it will also be
   used for intercomparison by the primary standard.
   The purpose of the batch control standard is to
   detect and assess time drift.

Between-set effect
   A statistically significant difference between
   2 or more sets of data.  If the
   sets came from different days, for example, then
   it would be said that "day has an effect",
   or that "there exists a between-day effect".

Block Plot
   A graphical data analysis technique to determine if
   a factor (e.g., cylinder) is statistically significant
   (yes/no), and if that significance holds
      locally:  for only some settings of other factors (e.g., day); or
      globally: for all settings of other factors (e.g., day).

Bracket the Target
   Primary standards chosen
   to be both above and below the nominal value
   of the SRM are said to "bracket the target".
   Experiment design questions in this regard are

      1) how many primary standards to choose?
      2) how far above and below?

Break
   Any interruption (coffee break, lunch break, end of day,
   power outage, etc.)
   which disrupts the physical flow of data during
   the chmeical analysis process.

Break Set (1, 2, 3, ...)
   The set of chemical measurements collected in the interim
   between 2 breaks.

..........C..........

CO
   Carbon monoxide.  This is a colorless, odorless gas.
   toxic?  stable?
   NIST provides CO SRMs as follows:
      xxx
      xxx

CO2
   Carbon dioxide.   colorless gas
   toxic?   stable?
   NIST provides CO2 SRMs as follows:
      xxx
      xxx

CU (see Certificate Uncertainty)

CV (see Certificate Value)

Calibration
   In general, this is the process whereby a
   high-accuracy estimate of some characteristic (such
   as concentration) of an artifact (such as a gas
   cylinder) is inferred by relating (through a
   calibration curve) the readings from a
   lower-accuracy device (such as a gas chromatograph)
   to the readings from higher-accuracy devices (via
   primary gas standards).

   For gas cylinder SRMs, this process has several steps:
      1) choice of the several primary standards to use
      2) choice of an appropriate calibration model/curve
         to fit when data is gathered;
      3) estimating the coefficients of the calibration model/curve
      4) translating (via the calibration curve) the
         low-accuracy reading to the corresponding
         high-accuracy reading.

Calibration Lot Control Standard
   This is the cylinder that is used for intercomparison
   during the calibration to the primary standard.
   In practice the calibration lot control standard is
   physically the same cylinder that is used as the lot
   control standard.

Calibration Curve
   A fitted mathematical model used for calibration purposes.
   A fitted model is one in which the theoretical coefficients
   of the model have been replaced by actual values as
   estimated from data.

Calibration Model
   A mathematical form (equation) which is used for calibration
   purposes.  The mathematical models most commonly
   used for calibration are

      1) constrained linear: straight line through origin;
      2) floating linear: straight kine with interecept anywhere;
      3) quadratic.

Certificate Uncertainty (CU)
   An SRM certificate has 2 entries:

      1) a certificate value (CV) which gives the best estimate
         for the concentration of the gas herein; and
      2) a certificate uncertainty (CU) for the certificate value.

   Statistical theory separates the uncertainty value
   into systematic and random components.
   ISO9000 recommendations separate the uncertainty value
   into "Type A" and "Type B" components.
   If a gas cylinders SRM carries the values 3.14 +- .07
   then the .07 is to be interpreted as follows:

      It is expected that the true (unknown) concentration
      within the cylinder is within the certificate uncertainty .07
      of the certificate value 3.14.--that is to say, the true value
      is expected to be between 3.07 and 3.21.

Certificate Value CV and CV(i)
   An SRM certificate has 2 entries:

      1) a certificate value (CV) which gives the best estimate
         for the concentration of the gas herein; and
      2) a certificate uncertainty (CU) for the certificate value.

   If a gas cylinders SRM carries the values 3.14 +- .07
   then the 3.14 is to be interpreted as follows:

      3.14 is our bstatistical est estimate of the true
      concentration in the cylinder.

      xxx

   The CV with no subscript implies global
   certification was done.  The CV with a subscript (as
   in CV(i)) implies individual certification was done.
   The certificate value is computed from the primary
   standard calibration.

   For global certification, the certificate value CV may be
   (depending on circumstances) computed as

      CV = Xhat
      CV = Xhat(i)
      CV = XhatCLCS * Rbar

   For individual certification, the certificate value CV(i) may be
   (depending on circumstances) computed as

      CV(i) = Xhat(i)
      CV(i) = XhatCLCS * Rbar(i)

Certification Experiment
   The in toto chemical analysis and statistical analysis protocol
   that is carried out in order to attach certificate
   values and uncertainties for an SRM.

Chemical Analysis
   Chemical analysis consists of all concentration measurement
   operations and calibration operations needed to obtain raw
   concentration readings.

Coded Cylinder ID (1 to 50)
   A positive integer (typically 1 to 50) which is attached
    to each cylinder in a lot of cylinders to be certified.
   The coded cylinder ID is useful in simplifying the statistical analysis.
   The coded cylinder ID is to be contrasted with the physical
   cylinder ID which is assigned by the cylinder manufacturer
   and is hard-stamped onto the cylinder identification plate.

Coded LCS ID
   The coded cylinder ID (between 1 and 50) of the
   lot control standard.

Comma-separated
   a collection of data such that each number on line (record)
   is separated from the next data point by a    comma.

Compare XhatCLCS for current time

Compare concentrations for current time

Computation Box

Concentrations for prior time.

Concentrations of LCS at time 1

Concentrations of LCS at time 2

Current Data

Cylinder & (Day or Break or LCS Set) Significant?

Cylinder Batch (1 to 4)

Cylinder ID
   A number or an alphanumeric mix which uniquely
   identifies a given cylinder.
   Cylinder IDs are of 2 types:
      1) coded cylinder id;
      2) physical cylinder id.

Cylinder Significant?
   Cylinder

Cylinder i

Click here to return to alphabetic list


Glossary D-E-F


..........D..........

DAN
   An acronym for "Data Analysis".
   Data analysis refers to all statistical procedures that
   are applied to the data after such data
   has in fact been collected.  In contrast, see DEX.

DEX
   An acronym for "Design of Experiment".
   DEX refers to those choices, strategies, and procedures
   that are executed prior to the collection of any data.
   The purpose of a good design of experiment is to assure that
      1) the SRM objectives of the experiment
         are attainable;
      2) the conclusions from the experiment are
         valid, uncontaminated, supportable, and repeatable;
      3) the experiment provide low-variability estimates
         of the desired parameters;
      4) the experiment be conducted in as
         efficient and inexpensive fashion as possible.

DEX Issues
DEX Recommendation
Data
Data Analysis
Data Analysis Protocol
Data File Format
Data File Recommendations
Data File Routinization
Data Manipulation
Data Reformatting
Data Separator
Dataplot
Date
Date
Day & Break or Day & LCS Set significant?
Day (1, 2, 3, ...)
Decay
Decision Point
Degrees of Freedom
Dependent Variable
Designed experiment
Developing the SRM
Df
Distinct Calibration Concentration
Distinct Calibration Concentrations
Distinct X
Distinct X's
Drift
Drift (= between-set effect)

..........E..........

EU = Expanded Uncertainty (global)
EU(i)= Expanded Uncertainty for Cylinder i
Effect
Enter RUPS = Relative uncertainty of the primary standard
Enter the coded ID number (1 to 50) for the
Estimated intermediate lot control standard values Lhat(i)
Excel
Experiment Design
Experimentalist/operator
Experimentalist/operator

..........F..........

F Test
Fabricating the SRM
Factor
Factor Significance
Factor Having an Effect
File
File Name
File name on line 1
Fit
Flowchart
Format

Click here to return to alphabetic list


Glossary G-H-I


..........G..........

GC
   Gas Chromatograph
Gas
Gas Chromatograph Port (1 to 16)

Gas ID
Global Certification
Global Decay
Global Drift
Global or Individual Certification
Globally Stable Gas
Graphical Data Analysis
Graphical Residual Analysis
Group
"Group"

..........H..........
H2S
   Hydrogen sulfide (an unstable gas)

Header
   The information provided at the beginning of a data file to describe
   its content.

Heterogeneous

Hidden Character
   Computer file information is typically of the alphabetic and
   numeric variety.  Some characters, however, may be invisible
   (hidden)--for example, the tab character, the escape character,
   etc. In particular, out of the 128 members of the ASCII character set,
   the first 32 of them are hidden (not visible).  For ease of I/O
   in the statistical analysis, it is recommended that no hidden
   characters be used in the data file.  In particular, the use of
   the tab character (since it cannot be seen but only its effect
   seen) is to be discoraged--a better choice would be to literally
   put in the blank spaces rather than the tab.

Homogeneous
   In general, homogeneous refers to consistency across elements.
   In particular for gas cylinders SRMs, homogeneous refers to
   a consistency across the cylinders of the batch to be certified.
   A gas cylinder batch is homogeneous if the concentration among
   the cylinders of this SRM batch are not significantly different.
   When the cylinders are homogeneous, the batch may be globally
   certified--that is a common gas concentration value is determined
   and that common value is assigned to all cylinders in the batch.

Homogeneous SRM
   A homogeneous SRM is one in which the batch of cylinders to be
   certified were homogeneous.  In such case, the cylinders are
   globally certified (with a common number) as opposed to
   individually certified (with cylinder-specific numbers).

..........I..........

Independent Variable X:  Primary Standard Concentration

Independent Variable
   The variable "x" in a regression model (also known as
   the regressor or predictor variable).

Individual Certification
   The certification method for an inhomogeneous gas
   cylinder SRM where certification are performed separately for each cylinder.

Inhomogeneous
   In general, inhomogeneous refers to inconsistency
   across elements.  In particular for gas cylinders
   SRMs, inhomogeneous refers to an inconsistency
   across the cylinders of the batch to be certified.

   The most common form of inconsistency is
   location-inconsistency: one or more of the cylinders
   out of the batch of 50 has a concentration which is
   higher (or lower) than the remaining cylinders.  A
   less common form of inconsistency is
   variation-inconsistency: one or more of the
   cylinders out of the batch of 50 has a distribution
   of multiple concentration readings which have a
   spread which is higher (or lower) than the remaining
   cylinders.

   If a batch of cylinders is said to be inhomogenous
   (with no explanation) then by default it is
   location-inconsistent.

Instability
   The quality or state of being unstable.

Instrument
   A device used to produce response (peak area, voltage, ...) for
   each cylinder.

Intercept
   The constant term in a model.

Interpreting Output

Click here to return to alphabetic list


Glossary J-K-L


..........J..........

..........K..........

..........L..........

LCS
   Lot control standard cylinder.

LCS Cylinder ID

LCS ID

LCS Set

Least Squares Regression
   The statistical method for determining the regression equation for
   which the sum of squares of the residuals is minimized.

Lhat
   Estimate of the variable "L".

Lhat(i)
   The estimated (usually by linear interpolation) ith intermediate
   response for lot control standard cylinder.

Linear Model
   The regression model of a line.

Linear-Through-Origin Model
   A linear model with zero intercept.

Local Drift

Local Decay

Lot Control Standard Cylinder
   The representative cylinder chosen from the lot
   of cylinders for each SRM to be intercompared to the other cylinders of the lot
   and to the primary gravimetric standards.

Click here to return to alphabetic list


Glossary M-N-O


..........M..........

Mean Response
  The expected value of a response (peak area, voltage, ...).

Mean Square Error
   The expected value of the square of an error.

Minimal Engineering Significant Difference

Model
  A mathematical description of an affair.

Multiple Calibration Experiments

..........N..........

n(i) = number of "groups" within cylinder i
n(i) = number of "groups" within day i
n(i) = number of ratios for cylinder i
nC = number of cylinders
nD = number of days
nG = number of "groups"
nmin = min{n(i)}
nR = number of ratios
NO
NTRM
Nominal Value

..........O..........

Omit data
Omit/correct data
Operating System
Optimize a process
Organics
Outlier
Outliers in the Lot Control Standard
Outliers in the Primary Standard Data
Outliers in the Test Cylinders
Outliers in the Ratios
Output Summary File

Click here to return to alphabetic list


Glossary P-Q-R


..........P..........

Parsimony
Parameters
PC 95
PC/Linix
PC/Unix
PC/Windows 95
PC/Windows NT
Pc/Windows 3.1
Physical cylinder ID
Platforms
Pool the 2 data sets.
Port Significant
Preprocess
Primary Standards Calibration Data
Primary Standards Calibration Data for Prior Time
Primary Standards Data
Prior Data
Process
Propagation of error
Propane
Protocol

..........Q..........

Quadratic : Xhat = ...
Quadratic : Xhat(i) = ...
Quadratic Model
Quadrature
Quantitative Data Analysis

..........R..........

REU
   Relative Expanded Uncertainty

RUC
   Rel. unc. of calibration  = RSD(Xhat(i))

RUC
   Rel. unc. of calibration = RSD(Xhat)

RUC
   Rel. unc. of calibration = RSD(XhatCLCS)

RUGM
   Rel. uncertainty of Rbar = SD(Rbar) / Rbar

RUGM
   Rela. unc. of grand mean = RSD(Rbar(i))

RUGM
   Rela. unc. of grand mean = RSD(Rbar)

RUGM(i)
   Rel. unc. of Rbar(i) = SD(Rbar(i)) / Rbar(i)

RUPS
   Relative Uncertainty of Primary Standard(s)

Ratio Statistics

Ratios
   Calculated ratio of response for sample cylinder,
   divided by measured or interpolated response for lot control
   standard cylinder.  Here "response" is a generic term for
   "voltage" or "peak area" or whatever the instrument provides
   as a measured value.

Raw Data
   Actual detailed data produced by the instrument,
   before ANY interpolation or averaging is done.

Rbar
   Average or a variable whose name is "R" (a ratio)
   Average of the nC Rbar(i)
   Average of the nD day averages
   Average of the nG Rbar(i) averages
   Average of the nR ratios

Rbar(i)
   Average of ratio data that belong to a defined
   subgroup labeled "i".  The index "i" may denote a day, or a
   "group" associated with a given pair of LCS readings, or the
   I.D. of a given cylinder, as determined by context.

Rbar(i,j)
   Average of ratio data that belong to a defined
   subgroup labeled "i,j".  In the case of two indexes, each
   index is used to label members of two different types of
   "group."  For example, "i" may denote "day" while "j"
   denotes "cylinder ID."  Or, depending on context, "i" could
   denote "cylinder ID" and "j" could denote "group associated
   with a pair of LCS readings."

Regression
   The statistical operation of fitting a line or
   curve by least squares.  The method used to calculate a
   slope, intercept, or the coefficients of a quadratic curve.

Replicate
   Repeated value, reading, or measurement (depending
   on context).

Residual Analysis
   The (usually graphical) process of checking
   the validity or appropriateness of a regression.  Outliers
   may be identified.  The decision may be made as to whether a
   straight line is adequate, as opposed to a quadratic.

Residual Standard Deviation
   Standard deviation representing
   the typical scatter of data around a regression line (or
   curve).

Response Variable Y
   Used to assign some part of the data as a
   variable named "Y".  This may be a voltage, or peak area, or
   a calculated ratio relative to the LCS.  Identifying this
   particular variable as the "response" and naming it "Y"
   implies that this variable will be used in a regression
   model as the dependent variable, represented as a linear (or
   quadratic) function of some other variable "x".

Routinization
   Making it possible to complete a statistical
   analysis by formally specifying an algorithm (or computer
   program or stepwise procedure) by which the necessary steps
   can be accomplished.

Run
   A generic term that refers to the (chemical or laboratory)
   operation required to produce one number in a table of data.
   The operational meaning of a "run" usually depends on the
   context of the data table under discussion.

Click here to return to alphabetic list


Glossary S-T-U


..........S..........

SAS
   Brand name of a large statistical software package.

SD(R)
   Standard deviation (actually standard uncertainty) of
   the quantity "R" inside the parentheses.

SD(Rbar)
   Standard deviation (actually standard uncertainty) of
   the quantity "Rbar" inside the parentheses.  The meaning of
   "Rbar" must be decided from context.

SD(Rbar(i))
   Standard deviation (actually standard uncertainty)
   of the quantity "Rbar(i)" inside the parentheses.  The
   meaning of "Rbar(i)" must be decided from context.
      SD of the nD day averages
      SD of the nG Rbar(i) averages
      SD(i) / sqrt(n(i))

SDRbar(i)
   SD of the n(i) Rbar(i,j) for cylinder i

SDpool
   "pooled" standard deviation obtained by combining (by
   weighted root mean square averaging) several independent
   standard deviations obtained from different subsets of the
   data.

Separator
   Tabs or commas or spaces used to separate numbers on
   the same line in a data file.  Spaces are preferred.

Significant Factor
   A factor (as Day, or Break, or LCS group,
   or Cylinder) that is determined to be a statistically
   significant contributor to the variability of the data.  For
   example, if the factor cylinder is significant, the it means
   that the cylinders differ in concentration (from each
   other).  If Day is significant, then there is a detectable
   amount of between-day variability that is not accounted for
   by the within-day variability.
      Significant Factors (including cylinder)
      Significant Factors (mode)

Significant Cylinder

Slope

SO2

Software

Spectroscopy

SRM

Scope

Stability
   Constancy.  For a cylinder, this refers to constancy
   of concentration over a longish period of time (6 months or
   a year).  For an instrument, this would refer to lack of
   drift during measurement.

Stable Gas

Standardized Reference Material

Standardized Data File Format

Standardize

Statistically Significant

Statistic Tools

Statistical Analysis

Sun
   A unix computer workstation used by many staff in the
   Statistical Engineering Division.

..........T..........

t TEST
Tab Character
Test Cylinder
The Calibration Lot Control Standards may drifted,
The process is complicated.
There is a drift in the concentration ??
Time 1
Time 2
Time effect?
Time of Day
Time of day not recorded: R(i) = Y(i) / Lhat(i)
Time of day recorded: R(t) = Y(t) / Lhat(t)
Turn-Around Time

..........U..........

Uncertainty Components
Unix
Unstable Gas

Click here to return to alphabetic list


Glossary V-W-X-Y-Z


..........V..........

Variable Data File Format
Vectorize
Vendor Data
Voltage

..........W..........

Web Page
What model to select?
Within-data-line separator = space(s)
Within-line separator
Write #1: Summary Information

..........X..........

Xhat = finverse(Rbar)
Xhat(i) = finverse(Rbar(i))
XhatCLCS = finverse(YbarCLCS)
XhatCLSC for prior time

..........Y..........

YbarCLCS = average of the Calibration Lot Control Standards
YhatCLCS

..........Z..........

Click here to return to FAQ list