QM0012 1 Explain the concept of Pareto Chart and Scatter Diagram.

Explanation of the concept of Pareto Chart Explanation of the concept of Scatter Diagram Also called: Pareto diagram Pareto anal!sis "ariations: #eighted Pareto chart comparati$e Pareto charts A Pareto chart is a %ar graph& 'he lengths of the %ars represent fre()enc! or cost *time or mone!+ and are arranged #ith longest %ars on the left and the shortest to the right& ,n this #a! the chart $is)all! depicts #hich sit)ations are more significant& 5 5 10

When to Use a Pareto Chart

-hen anal!.ing data a%o)t the fre()enc! of pro%lems or ca)ses in a process& -hen there are man! pro%lems or ca)ses and !o) #ant to foc)s on the most significant& -hen anal!.ing %road ca)ses %! loo/ing at their specific components& -hen comm)nicating #ith others a%o)t !o)r data&

Pareto Chart Procedure

1& Decide #hat categories !o) #ill )se to gro)p items& 2& Decide #hat meas)rement is appropriate& Common meas)rements are fre()enc! ()antit! cost and time& 0& Decide #hat period of time the Pareto chart #ill co$er: 1ne #or/ c!cle2 1ne f)ll da!2 A #ee/2 3& Collect the data recording the categor! each time& *1r assem%le data that alread! exist&+ 5& S)%total the meas)rements for each categor!& 4& Determine the appropriate scale for the meas)rements !o) ha$e collected& 'he maxim)m $al)e #ill %e the largest s)%total from step 5& *,f !o) #ill do optional steps 5 and 6 %elo# the maxim)m $al)e #ill %e the s)m of all s)%totals from step 5&+ Mar/ the scale on the left side of the chart& 7& Constr)ct and la%el %ars for each categor!& Place the tallest at the far left then the next tallest to its right and so on& ,f there are man! categories #ith small meas)rements the! can %e gro)ped as 8other&9 Steps 5 and 6 are optional %)t are )sef)l for anal!sis and comm)nication&

5& Calc)late the percentage for each categor!: the s)%total for that categor! di$ided %! the total for all categories& Dra# a right $ertical axis and la%el it #ith percentages& :e s)re the t#o scales match: ;or example the left meas)rement that corresponds to one<half sho)ld %e exactl! opposite 50= on the right scale& 6& Calc)late and dra# c)m)lati$e s)ms: Add the s)%totals for the first and second categories and place a dot a%o$e the second %ar indicating that s)m& 'o that s)m add the s)%total for the third categor! and place a dot a%o$e the third %ar for that ne# s)m& Contin)e the process for all the %ars& Connect the dots starting at the top of the first %ar& 'he last dot sho)ld reach 100 percent on the right scale&

Pareto Chart Examples

Example >1 sho#s ho# man! c)stomer complaints #ere recei$ed in each of fi$e categories& Example >2 ta/es the largest categor! 8doc)ments 9 from Example >1 %rea/s it do#n into six categories of doc)ment<related complaints and sho#s c)m)lati$e $al)es& ,f all complaints ca)se e()al distress to the c)stomer #or/ing on eliminating doc)ment< related complaints #o)ld ha$e the most impact and of those #or/ing on ()alit! certificates sho)ld %e most fr)itf)l&

Example >1

a. Explain Poisson distribution. Give any two examples of Poisson distribution. b. If ! of electric bulbs manufactured by a certain company are defective" find the probability that in a sample of ## bulbs i$ less than defective.&e ) #.#*+%, Explanation of Poisson Distri%)tion Examples of Poisson Distri%)tion Calc)lation?Sol)tion to the pro%lem
'(

bulbs" ii$ more than % bulbs are 3 1 5 10

The outcome of an experiment need not be a number, for example, the outcome when a coin is tossed can be 'heads' or 'tails'. However, we often want to represent outcomes as numbers. A random variable is a function that associates a unique numerical value with every outcome of an experiment. The value of the random variable will vary from trial to trial as the experiment is repeated. There are two types of random variable - discrete and continuous. A random variable has either an associated probability distribution (discrete random variable) or probability density function (continuous random variable). Examples 1. A coin is tossed ten times. The random variable X is the number of tails that are noted. X can only take the values 0, 1, ..., 10, so X is a discrete random variable.

2. A light bulb is burned until it burns out. The random variable Y is its lifetime in hours. Y can take any positive real value, so Y is a continuous random variable.

ExpectedValue The expected value (or population mean) of a random variable indicates its average or central value. It is a useful summary value (a number) of the variable's distribution. Stating the expected value gives a general impression of the behaviour of some random variable without giving full details of its probability distribution (if it is discrete) or its probability density function (if it is continuous). Two random variables with the same expected value can have very different distributions. There are other useful descriptive measures which affect the shape of the distribution, for example variance. The expected value of a random variable X is symbolised by E(X) or . If X is a discrete random variable with possible values x1, x2, x3, ..., xn, and p(xi) denotes P(X = xi), then the expected value of X is defined by: where the elements are summed over all values of the random variable X. If X is a continuous random variable with probability density function f(x), then the expected value of X is defined by: Example Discrete case : When a die is thrown, each of the possible faces 1, 2, 3, 4, 5, 6 (the xi's) has a probability of 1/6 (the p(xi)'s) of showing. The expected value of the face showing is therefore: = E(X) = (1 x 1/6) + (2 x 1/6) + (3 x 1/6) + (4 x 1/6) + (5 x 1/6) + (6 x 1/6) = 3.5 Notice that, in this case, E(X) is 3.5, which is not a possible value of X. See also sample mean.

Variance The (population) variance of a random variable is a non-negative number which gives an idea of how widely spread the values of the random variable are likely to be; the larger the variance, the more scattered the observations on average. Stating the variance gives an impression of how closely concentrated round the expected value the distribution is; it is a measure of the 'spread' of a distribution about its average value. Variance is symbolised by V(X) or Var(X) or The variance of the random variable X is defined to be: where E(X) is the expected value of the random variable X.
0

Explain the procedure generally followed in testing of a hypothesis. -rite a detailed note on .ne' tailed and /wo'tailed tests. Proced)re follo#ed in testing h!pothesis Explanation of 1ne tailed test Explanation of '#o tailed test 5 2&5 2&5 10

S)ppose #e ha$e a n)ll h!pothesis @0 and an alternati$e h!pothesis @1& -e consider the distri%)tion gi$en %! the n)ll h!pothesis and perform a test to determine #hether or not the n)ll h!pothesis sho)ld %e reAected in fa$o)r of the alternati$e h!pothesis& 'here are t#o different t!pes of tests that can %e performed& A one-tailed test loo/s for an increase or decrease in the parameter #hereas a two-tailed test loo/s for an! change in the parameter *#hich can %e an! change< increase or decrease+& -e can perform the test at an! le$el *)s)all! 1= 5= or 10=+& ;or example performing the test at a 5= le$el means that there is a 5= chance of #rongl! reAecting @0& ,f #e perform the test at the 5= le$el and decide to reAect the n)ll h!pothesis #e sa! Bthere is significant e$idence at the 5= le$el to s)ggest the h!pothesis is falseB& One-Tailed Test

-e choose a critical region& ,n a one<tailed test the critical region #ill ha$e A)st one part *the red area %elo#+& ,f o)r sample $al)e lies in this region #e reAect the n)ll h!pothesis in fa$o)r of the alternati$e& S)ppose #e are loo/ing for a definite decrease& 'hen the critical region #ill %e to the left& Cote ho#e$er that in the one<tailed test the $al)e of the parameter can %e as high as !o) li/e&

Example S)ppose #e are gi$en that D has a Poisson distri%)tion and #e #ant to carr! o)t a h!pothesis test on the mean %ased )pon a sample o%ser$ation of 0& S)ppose the h!potheses are: @0: E 6 @1: F 6 -e #ant to test if it is Breasona%leB for the o%ser$ed $al)e of 0 to ha$e come from a Poisson distri%)tion #ith parameter 6& So #hat is the pro%a%ilit! that a $al)e as lo# as 0 has come from a Po*6+2 P*D F 0+ E 0&0212 *this has come from a Poisson ta%le+ 'he pro%a%ilit! is less than 0&05 so there is less than a 5= chance that the $al)e has come from a Poisson*0+ distri%)tion& -e therefore reAect the n)ll h!pothesis in fa$o)r of the alternati$e at the 5= le$el& @o#e$er the pro%a%ilit! is greater than 0&01 so #e #o)ld not reAect the n)ll h!pothesis in fa$o)r of the alternati$e at the 1= le$el& < See more at: http:??###&mathsre$ision&net?ad$anced<le$el<maths< re$ision?statistics?one<and<t#o<tailed<tests>sthash&#)gA5lG'&dp)f
3

-hat is analysis of variance0 State all the assumptions involved in analysis of variance

techni1ue. Explain the structure for .ne way analysis of variance or one way classification. Meaning of anal!sis of $ariance Ass)mptions Str)ct)re of 1ne #a! anal!sis of $ariance 2 0 5
10

Analysis of variance *ANO A+ is a collection of statistical models )sed to anal!.e the differences %et#een gro)p means and their associated proced)res *s)ch as B$ariationB among and %et#een gro)ps+& ,n AC1"A setting the o%ser$ed $ariance in a partic)lar $aria%le is partitioned into components attri%)ta%le to different so)rces of $ariation& ,n its simplest form AC1"A pro$ides a statistical test of #hether or not the means of se$eral gro)ps are e()al and therefore generali.es t<test to more than t#o gro)ps& Doing m)ltiple t#o<sample t<tests #o)ld res)lt in an increased chance of committing a t!pe , error& ;or this reason AC1"As are )sef)l in comparing *testing+ three or more means *gro)ps or $aria%les+ for statistical significance& 'he anal!sis of $ariance can %e )sed as an explorator! tool to explain o%ser$ations& A dog sho# pro$ides an example& A dog sho# is not a random sampling of the %reed& ,t is t!picall! limited to dogs that are male ad)lt p)re<%red and exemplar!& A histogram of dog #eights from a sho# might pla)si%l! %e rather complex li/e the !ello#<orange distri%)tion sho#n in the ill)strations& An attempt to explain the #eight distri%)tion %! di$iding the dog pop)lation into gro)ps *!o)ng $s old+*short<haired $s long<haired+ #o)ld pro%a%l! %e a fail)re *no fit at all+& 'he gro)ps *sho#n in %l)e+ ha$e a large $ariance and the means are $er! close& An attempt to explain the #eight distri%)tion %! *pet $s #or/ing %reed+*less athletic $s more athletic+ #o)ld pro%a%l! %e some#hat more s)ccessf)l *fair fit+& 'he hea$iest sho# dogs are li/el! to %e %ig strong #or/ing %reeds& An attempt to explain #eight %! %reed is li/el! to prod)ce a $er! good fit& All Chih)ah)as are light and all St :ernards are hea$!& 'he difference in #eights %et#een Setters and Pointers does not A)stif! separate %reeds& 'he anal!sis of $ariance pro$ides the formal tools to A)stif! these int)iti$e A)dgments& A common )se of the method is the anal!sis of experimental data or the de$elopment of models& 'he method has some ad$antages o$er correlation: not all of the data m)st %e n)meric and one res)lt of the method is a A)dgment in the confidence in an explanator! relationship& AC1"A is a partic)lar form of statistical h!pothesis testing hea$il! )sed in the anal!sis of experimental data& A statistical h!pothesis test is a method of ma/ing decisions )sing data& A test res)lt *calc)lated from the n)ll h!pothesis and the sample+ is called statisticall! significant if it is deemed )nli/el! to ha$e occ)rred %! chance assuming the truth of the null hypothesis& A statisticall! significant res)lt *#hen a pro%a%ilit! *p<$al)e+ is less than a threshold *significance le$el++ A)stifies the reAection of the n)ll h!pothesis& ,n the t!pical application of AC1"A the n)ll h!pothesis is that all gro)ps are simpl! random samples of the same pop)lation& 'his implies that all treatments ha$e the same

effect *perhaps none+& HeAecting the n)ll h!pothesis implies that different treatments res)lt in altered effects& :! constr)ction h!pothesis testing limits the rate of '!pe , errors *false positi$es leading to false scientific claims+ to a significance le$el& Experimenters also #ish to limit '!pe ,, errors *false negati$es res)lting in missed scientific disco$eries+& 'he '!pe ,, error rate is a f)nction of se$eral things incl)ding sample si.e *positi$el! correlated #ith experiment cost+ significance le$el *#hen the standard of proof is high the chances of o$erloo/ing a disco$er! are also high+ and effect si.e *#hen the effect is o%$io)s to the cas)al o%ser$er '!pe ,, error rates are lo#+& 'he terminolog! of AC1"A is largel! from the statistical design of experiments& 'he experimenter adA)sts factors and meas)res responses in an attempt to determine an effect& ;actors are assigned to experimental )nits %! a com%ination of randomi.ation and %loc/ing to ens)re the $alidit! of the res)lts& :linding /eeps the #eighing impartial& Hesponses sho# a $aria%ilit! that is partiall! the res)lt of the effect and is partiall! random error& AC1"A is the s!nthesis of se$eral ideas and it is )sed for m)ltiple p)rposes& As a conse()ence it is diffic)lt to define concisel! or precisel!& BClassical AC1"A for %alanced data does three things at once: 1& As explorator! data anal!sis an AC1"A is an organi.ation of an additi$e data decomposition and its s)ms of s()ares indicate the $ariance of each component of the decomposition *or e()i$alentl! each set of terms of a linear model+& 2& Comparisons of mean s()ares along #ith ;<tests &&& allo# testing of a nested se()ence of models& 0& Closel! related to the AC1"A is a linear model fit #ith coefficient estimates and standard errors&B ,n short AC1"A is a statistical tool )sed in se$eral #a!s to de$elop and confirm an explanation for the o%ser$ed data& Additionall!: 3& ,t is comp)tationall! elegant and relati$el! ro%)st against $iolations of its ass)mptions& 5& AC1"A pro$ides ind)strial strength *m)ltiple sample comparison+ statistical anal!sis& 4& ,t has %een adapted to the anal!sis of a $ariet! of experimental designs& As a res)lt: AC1"A Bhas long enAo!ed the stat)s of %eing the most used *some #o)ld sa! a%)sed+ statistical techni()e in ps!chological research&B AC1"A Bis pro%a%l! the most useful techni()e in the field of statistical inference&B

AC1"A is diffic)lt to teach partic)larl! for complex experiments #ith split<plot designs %eing notorio)s&I3J ,n some cases the proper application of the method is %est determined %! pro%lem pattern recognition follo#ed %! the cons)ltation of a classic a)thoritati$e test&I5J

Design-of-experiments terms
*Condensed from the C,S' Engineering Statistics hand%oo/: Section 5&7& A Klossar! of D1E 'erminolog!&+I4J :alanced design An experimental design #here all cells *i&e& treatment com%inations+ ha$e the same n)m%er of o%ser$ations& :loc/ing A sched)le for cond)cting treatment com%inations in an experimental st)d! s)ch that an! effects on the experimental res)lts d)e to a /no#n change in ra# materials operators machines etc& %ecome concentrated in the le$els of the %loc/ing $aria%le& 'he reason for %loc/ing is to isolate a s!stematic effect and pre$ent it from o%sc)ring the main effects& :loc/ing is achie$ed %! restricting randomi.ation&
5

Explain the different types of attribute control charts. Distinguish between control charts for variables and control charts for attributes. Meaning of attri%)te control charts '!pes of attri%)te control charts Differences 2 3 3
10

Control charts also /no#n as !hewhart charts *after -alter A& She#hart+ or process"ehavior charts in statistical process control are tools )sed to determine if a man)fact)ring or %)siness process is in a state of statistical control& ,f anal!sis of the control chart indicates that the process is c)rrentl! )nder control *i&e& is sta%le #ith $ariation onl! coming from so)rces common to the process+ then no corrections or changes to process control parameters are needed or desired& ,n addition data from the process can %e )sed to predict the f)t)re performance of the process& ,f the chart indicates that the monitored process is not in control anal!sis of the chart can help determine the so)rces of $ariation as this #ill res)lt in degraded process performance& I1J A process that is sta%le %)t operating o)tside of desired *specification+ limits *e&g& scrap rates ma! %e in statistical control %)t a%o$e desired limits+ needs to %e impro$ed thro)gh a deli%erate effort to )nderstand the ca)ses of c)rrent performance and f)ndamentall! impro$e the process&

'he control chart is one of the se$en %asic tools of ()alit! control& '!picall! control charts are )sed for time<series data tho)gh the! can %e )sed for data that ha$e logical compara%ilit! *i&e& !o) #ant to compare samples that #ere ta/en all at the same time or the performance of different indi$id)als+ ho#e$er the t!pe of chart )sed to do this re()ires consideration&

#istory
'he control chart #as in$ented %! -alter A& She#hart #hile #or/ing for :ell La%s in the 1620s& 'he compan!Ms engineers had %een see/ing to impro$e the relia%ilit! of their telephon! transmission s!stems& :eca)se amplifiers and other e()ipment had to %e %)ried )ndergro)nd there #as a %)siness need to red)ce the fre()enc! of fail)res and repairs& :! 1620 the engineers had alread! reali.ed the importance of red)cing $ariation in a man)fact)ring process& Moreo$er the! had reali.ed that contin)al process<adA)stment in reaction to non<conformance act)all! increased $ariation and degraded ()alit!& She#hart framed the pro%lem in terms of Common< and special<ca)ses of $ariation and on Ma! 14 1623 #rote an internal memo introd)cing the control chart as a tool for disting)ishing %et#een the t#o& She#hartMs %oss Keorge Ed#ards recalled: BDr& She#hart prepared a little memorand)m onl! a%o)t a page in length& A%o)t a third of that page #as gi$en o$er to a simple diagram #hich #e #o)ld all recogni.e toda! as a schematic control chart& 'hat diagram and the short text #hich preceded and follo#ed it set forth all of the essential principles and considerations #hich are in$ol$ed in #hat #e /no# toda! as process ()alit! control&BI5J She#hart stressed that %ringing a prod)ction process into a state of statistical control #here there is onl! common<ca)se $ariation and /eeping it in control is necessar! to predict f)t)re o)tp)t and to manage a process economicall!& She#hart created the %asis for the control chart and the concept of a state of statistical control %! caref)ll! designed experiments& -hile She#hart dre# from p)re mathematical statistical theories he )nderstood data from ph!sical processes t!picall! prod)ce a Bnormal distri%)tion c)r$eB *a Ka)ssian distri%)tion also commonl! referred to as a B%ell c)r$eB+& @e disco$ered that o%ser$ed $ariation in man)fact)ring data did not al#a!s %eha$e the same #a! as data in nat)re *:ro#nian motion of particles+& She#hart concl)ded that #hile e$er! process displa!s $ariation some processes displa! controlled $ariation that is nat)ral to the process #hile others displa! )ncontrolled $ariation that is not present in the process ca)sal s!stem at all times& ,n 1623 or 1625 She#hartMs inno$ation came to the attention of -& Ed#ards Deming then #or/ing at the @a#thorne facilit!& Deming later #or/ed at the Nnited States Department of Agric)lt)re and %ecame the mathematical ad$isor to the Nnited States Cens)s :)rea)& 1$er the next half a cent)r! Deming %ecame the foremost champion and proponent of She#hartMs #or/& After the defeat of Oapan at the close of -orld -ar ,, Deming ser$ed as statistical cons)ltant to the S)preme Commander for the Allied Po#ers& @is ens)ing in$ol$ement in Oapanese life and long career as an ind)strial cons)ltant there spread She#hartMs thin/ing and the )se of the control chart #idel! in Oapanese man)fact)ring ind)str! thro)gho)t the 1650s and 1640s&

Chart details
A control chart consists of:

Points representing a statistic *e&g& a mean range proportion+ of meas)rements of a ()alit! characteristic in samples ta/en from the process at different times Ithe dataJ 'he mean of this statistic )sing all the samples is calc)lated *e&g& the mean of the means mean of the ranges mean of the proportions+ A centre line is dra#n at the $al)e of the mean of the statistic 'he standard error *e&g& standard de$iation?s(rt*n+ for the mean+ of the statistic is also calc)lated )sing all the samples Npper and lo#er control limits *sometimes called Bnat)ral process limitsB+ that indicate the threshold at #hich the process o)tp)t is considered statisticall! M)nli/el!M and are dra#n t!picall! at 0 standard errors from the centre line

'he chart ma! ha$e other optional feat)res incl)ding:

Npper and lo#er #arning or control limits dra#n as separate lines t!picall! t#o standard errors a%o$e and %elo# the centre line Di$ision into .ones #ith the addition of r)les go$erning fre()encies of o%ser$ations in each .one Annotation #ith e$ents of interest as determined %! the Q)alit! Engineer in charge of the processMs ()alit!

Explain the methodology for Statistical Process Control implementation 2SPC$. -hat are the benefits derived from SPC0
Methodolog! of SPC :enefits 4 3 10

'he concepts of Statistical Process Control *SPC+ #ere initiall! de$eloped %! $r% Walter !hewhart of :ell La%oratories in the 1620Ms and #ere expanded )pon %! $r% W% Edwards $emin& #ho introd)ced SPC to Oapanese ind)str! after --,,& After earl! s)ccessf)l adoption %! Oapanese firms Statistical Process Control has no# %een

incorporated %! organi.ations aro)nd the #orld as a primar! tool to impro$e prod)ct ()alit! %! red)cing process $ariation& Dr& She#hart identified t#o so)rces of process $ariation: Chance $ariation that is inherent in process and sta%le o$er time and Assi&na"le or Uncontrolled $ariation #hich is )nsta%le o$er time < the res)lt of specific e$ents o)tside the s!stem& Dr& Deming rela%eled chance $ariation as Common Cause $ariation and assigna%le $ariation as !pecial Cause $ariation& :ased on experience #ith man! t!pes of process data and s)pported %! the la#s of statistics and pro%a%ilit! Dr& She#hart de$ised control charts )sed to plot data o$er time and identif! %oth Common Ca)se $ariation and Special Ca)se $ariation& This tutorial provides a "rief conceptual "ac'&round to the practice of !PC( as well as the necessary formulas and techni)ues to apply it%

Process aria"ility
,f !o) ha$e re$ie#ed the disc)ssion of fre()enc! distri%)tions in the @istogram mod)le !o) #ill recall that man! histograms #ill approximate a Cormal Distri%)tion as sho#n %elo# *please note that control charts do not re)uire normally distri"uted data in order to wor' < the! #ill #or/ #ith an! process distri%)tion < #e )se a normal distri%)tion in this example for ease of representation+:

,n order to #or/ #ith an! distri%)tion it is important to ha$e a meas)re of the data dispersion or spread& 'his can %e expressed %! the range *highest less lo#est+ %)t is %etter capt)red %! the standard de$iation *sigma+& 'he standard de$iation can %e easil! calc)lated from a gro)p of n)m%ers )sing man! calc)lators or a spreadsheet or statistics program&

Example

Why *s $ispersion !o *mportant+

1ften #e foc)s on a$erage $al)es %)t )nderstanding dispersion is critical to the management of ind)strial processes& Consider t#o examples:

,f !o) p)t one foot in a %)c/et of ice #ater *00 degrees ;+ and one foot in a %)c/et of scalding #ater *127 degrees ;+ on a$erage !o)Mll feel fine *50 degrees ;+ %)t !o) #onMt act)all! %e $er! comforta%leP ,f !o) are as/ed to #al/ thro)gh a ri$er and are told that the a$erage #ater depth is 0 feet !o) might #ant more information& ,f !o) are then told that the range is from .ero to 15 feet !o) might #ant to re<e$al)ate the trip&

,ore!team #int- Anal!sis of a$erages sho)ld al#a!s %e accompanied %! anal!sis of the $aria%ilit!P

Control .imits
Statistical ta%les ha$e %een de$eloped for $ario)s t!pes of distri%)tions that ()antif! the area )nder the c)r$e for a gi$en n)m%er of standard de$iations from the mean *the normal distri%)tion is sho#n in this example+& 'hese can %e )sed as pro%a%ilit! ta%les to calc)late the odds that a gi$en $al)e *meas)rement+ is part of the same gro)p of data )sed to constr)ct the histogram& She#hart fo)nd that control limits placed at three standard de$iations from the mean in either direction pro$ide an economical tradeoff %et#een the ris/ of reacting to a false signal and the ris/ of not reacting to a tr)e signal < regardless the shape of the )nderl!ing process distri%)tion& ,f the process has a normal distri%)tion 66&7= of the pop)lation is capt)red %! the c)r$e at three standard de$iations from the mean& Stated another #a! there is onl! a 1<66&7= or 0&0= chance of finding a $al)e %e!ond 0 standard de$iations& 'herefore a

meas)rement $al)e %e!ond 0 standard de$iations indicates that the process has either shifted or %ecome )nsta%le *more $aria%ilit!+&