 Hello everyone, welcome to the session 4 of Quality Control and Improvement Using Minitab. I am Professor Indrajit Mukherjee, Chair of HMA, the School of Management, IIT Bombay. So today our agenda will be to understand how to visualize the data basically. So last session what we have done is that we have discussed about Canon models, QFD and we have tried to link between CTQs and voice of the customers like that. We are interested totally now on CTQs. So quality of conformance is important for us, control and conformance and improvement of the CTQ is our agenda. So quality is all about improving CTQs. So either I control or I improve the CTQs like that. So we will discuss more about that. So last time we have talked about some aspects of CTQs and we have used a diagram to represent that one. So this is the diagram that we are using that there will be inputs and there will be control variables which can be monitored and can be changed. There will be some noise factors which we do not have any control as such, but we want to minimize their effects and we want the quality output CTQ to be close to the target that is defined by the designer and with minimum variability we want, we want to measure that one. What is happening? What is the outcome like that? So whenever we measure we have some data on CTQs and then we are going to visualize the data or CTQs. So what is the status of the data? So for that we have shown that strength is CTQ then we can draw a histogram over here and histograms are like this. So there will be frequency axis on one direction and there will be strength in one direction. So this will give you some idea where is the mean of the data set and what is the spread of the data set basically? What is the spread of the data set? So mean of the data set of a sample data set we can express by using a symbol rotation that x bar and the spread of the data set can be expressed as standard deviation of the data set and I am assuming that basic idea of mean and standard deviation is already you have and so we will go ahead with that idea that I understand mean and standard deviation. So but in quality concept what mean and standard deviation means basically that will be our discussion on session 4 and what are the what are the techniques that is used to minimize the mean variance and also bring the mean to targets that that is our overall objective. So let us try to understand one important thing in quality which is known as accuracy and precisions like that. So when we talk about accuracy we are talking about the mean of the data set or center of gravity of the data sets like that or we can we can we can think of the location of the data sets basically ok. So CG of the data set center of gravity of the data set central tendency of the data set you can you can give different names like that ok. So one is mean how close is that to the target values and what is what is precision basically. Whenever I am talking about precision we are talking about variance over here. So variance is expressed as a square like that and if you make a square root of that that will give you the sample standard deviations ok. So so this is a bullseye kind of scenario so where target value is already defined so you have to if you are hitting this point you are getting the maximum score like that ok. So but people generally tries to miss the miss the target so they will there will be variations from person to person and in process also like that. So there will be inputs and there will be process settings, there will be noises and that will create variations in the outcome that is CTQ outcomes like that ok. So there can be different types of variation there can be different types of errors in either in accuracy part or in precision part like that ok. So if you see the diagram over here the one axis is accuracy over here and one axis is precision over here that is shown on the x axis like that ok. So whenever my accuracy and precision both are high in that case I will hit the target and with minimum variability that is the diagram that is the most best scenario we can we can we can think of best scenario we can think of. So what is what scenario over here what scenario is not accurate not precise so accuracy is also low and precision is also low like that. So this is what scenario we can think of over here ok. So here what you can see is that I am hitting the target with minimum variation over here variability is very less over here ok. And over here the other diagram what is case is not accurate not even variability you can see much larger variability what you can see over here and off target is I am not hitting the target so overall average is also off centered like that. So it will come to be off centered like that. So variation is also high so that is the worst case. So what can happen in between is that so I have precision I have high end precision but I am off centered like that. So this kind of scenario can be improved if my process locations can be shifted that means the setting can be changed so that only my location is shifted the outcome CTQ the average outcome of the CTQ is near to the target like that. So now I am precise in the outcomes of the CTQs but I am not accurate basically so I can change that one. So this kind of correction is quite easy in manufacturing process. So in this case this is high precision but low accuracy so I have to improve the accuracy over here. And here what happens is that the average comes out to be near the target but variability high this is also not a favorable situation for us. So in this case what I have to do is that I have to reduce the variability like that. So in any processes our objective in quality is to increase the accuracy and increase the precision that means mean should be in the target value and target means what designer has defined the target value or customer has defined the target value like that. So I want to be in a process precise and accurate both the things are important. So both are important. So let us take an example let us give an example like Piano Manufacturing. So this was an evidence case where a well known company was manufacturing Piano for high end customers like that. So they wanted to launch a new model which is at the low end like that not on high end products like that. So they want to something which is in between products which is which is not so high quality like that. So they want to sell those products also. So that category of products they want to sell like that. So and they have a competitor also because competitors are giving them that pressure they are feeling that those companies are and those companies agenda was to confirm to the products specification like that. But this Piano Manufacturing what they are doing they are customizing the products like that then they are selling to the high end customers or who are knowledgeable customers like that. And for every customer Piano sound quality is very important like that. So and the performance of the Piano across checks inspection was there R and D was there to improve that one they have a high end quality like that. So in that case what comes out of that there are two definitions what comes out of that one is conformance to quality and one is performance that means what what we have defined is quality of design and quality of performance basically. When it goes to the end users how it is performing basically and conformance means I have the design quality of design and then I am manufacturing that one. So the competitors are they are having they are focused on conformance basically. So I am within the specification. So in this case they are manufacturing some products which is within the specification and they are doing it consistently like that. And this is not to the high end customers. So customer segment is different over here. So in this case what was happening they are confirming but what happens is that in Piano manufacturing thousands of components goes into manufacturing of the pianos like that. So in that case what happens is that there will be interplay between the parts like that. So if the parts are only confirming to specification one is at the extreme end of the specification this can be upper specification this can be lower specification. So one component may be manufactured at this range and some other component. So component one like this some other component may be manufactured at this zone at the lower end like that. So different components so when this component C2 and C1 when you when you assemble both those parts what will happen is that all are in the extreme of the specifications like that. So the we will get more noises over here. So in this case what will happen is that overall performance will deteriorate like that. So there will be interplay between this component one and component two. So but what is required is basically component one should also fulfill targets it is having a target one like that it will having a target two. So if everything is on target component one and component two what will happen is that the overall performance of the process overall performance of the products will be will be the best. So there will be little interplay little noises like that. So ultimately the performance will improve. So any company's objective is to any component that is coming out of the process should be accurate and precise like that. So if I only confirm that will not help basically to bring out high performance like that. So conformance means within specification and they are consistently doing that. So consistency in that doing conformance does not ensure performance like that. So whenever you have to perform so it has to be high end performance in quality. So in that case every component is important to me every component is important to me. So both accuracy and precision of every parts is important basically we cannot ignore any of the parts and say conformance. So that cold post mentality we cannot we cannot believe in that. So only within specification will not help. So every product should be manufactured on the target and the accuracy of the product and precision of the product should be very high like that. Then only the performance will come or on-road performance of a car will happen only if each of the components are manufactured with accurate with ah with near to the targets and also the precision of the or variability of the components that are coming out of the process is very less very very less like that ok. So that is the overall idea. So in quality also accuracy and precision is very important accuracy and precision both are important to us ok. So you you may have different types of data and measure of accuracy can be different measure of precision can be different like that. But anyhow we are assuming that CTQ can be measured and that way we can improve the accuracy and we can improve the precision also ok. So whenever the CTQ is we are measuring the CTQs what will happen there will be variation in the CTQs like that. So there can be different types of variability what what generally quality persons are interested into. So one of the one of the variability is known as common cause variability or sample to sample variability. So this kind of variability is very very much it is very difficult to reduce this kind of variabilities like that and this is when the input condition changes it will it will influence whether whether your setting condition is same. But because the inputs has changed certain amount of inputs has changed or noise condition has changed. So that will impact the outcomes or CTQs like that and there will be variation like that. So this is small small variation what happens in the process that is common cause variability. So we initially we we do not consider of this kind of variabilities like that later on we will consider this kind of variabilities. But then there are other kinds of variability which is which can be easily handled or which can be we can deal with those things that is that is known as special cause variability or abnormal variability. You go to a process and what you see is that abnormal vibration is happening on the machines or the operator has completely changed like that there were skilled operators now there is new operator like that operator does not know how to set the machine what will happen is that the process behavior will change centering will change like that. Similarly you can catch hold of those things that this as this is going wrong basically. So abnormal variability is easy to detect and then you have to take action over there to reduce that kind of variability like that. Although it is time consuming you need to eliminate basically abnormal variations like that this are known as special cause variability and there are quality tools to address this one and special cause we can think of abnormal variations or assignable cause variations assignable cause special cause variations like that we can we can we can name that in different ways like that there are different names for this type of variability. So so these are very very peculiar type of variations. So suddenly behavior abruptly changes like that so ok. So this can be we we use a special kind of techniques in quality which is known as statistical process control to identify which is normal and which is abnormal like that. So there will be common variations within the circles that you are seeing. So these variations are common variability we will try to decrease that one after means in using different techniques like that. But currently I want to differentiate between let us say common variability and special cause. So I have a yellow yellow zone what you see over here. So I can define a zone like that whenever it falls beyond this blue zone over here we say that there is something going wrong in the process that means suddenly abrupt something has happened in the process like that. So that is a special cause and immediately the process will be stopped diagonalized like that and we will take some corrective action like that. So so so alarm will we will set an alarm like that. So it is like signaling system also we can think of. So whenever it is yellow we should be very precautious like that we should say take some corrective actions like that. But red zone what does it indicate maybe out of spec out of specifications basically ok. So in yellow zone over here we we we get concerned and we we try to eliminate why this is happening like that. And if it is red zone product maybe discarded all together like that it has gone out of specification what is defined by the design and like that. So there will be common cause variation there will be special cause variation which we want to block like that and if it goes beyond certain extent the product may be rejected like that rejected like that. So one is common cause variations which generally happens in a process for sample to sample variation we can think of one is special cause variation that are that are detectable. So in that case and that can be eliminated if I if I can eliminate recurrence of those time those kind of causes like that some amount of variability can be reduced over here. And the third one is systematic variation that means I intentionally induce variability in the process. So that is known as experimentation in the process. What we try to do over here is that basically there is a process over here and there will be what diagram what we have shown is y is coming out of the process there will be x variables over here. This is the CTQ that we are measuring over here. So what we try to do by experimentation is that we try to build a function between y and x over here and then try to determine what is the optimal condition of x that will give me the best y or y hitting the target value with minimum variation that s is over here and the average value is coming to look close to the target values over here. So what should be the condition of x? But first what we need is that what is the functional relationship? If I get a function then I can optimize that function like that. So systematic cause or intentional variation is induced to understand the behavior and understand the relationship between y and x and so that we can model that one and we can optimize that one. So that is known as design of experiment part. So we have a control part which is known as statistical process control. We have an experimentation part and control part together over here which there is a we are doing experimentation. We are setting the process. We are setting the standard operating practice and we are controlling that also. So that is the ultimate thing that in quality people tries to achieve like that ok. So and this systematic experimentation how it will help is that it will even reduce the common cause variability over here. So that you see a common cause variations can with the influence even even with the influence of common cause variability the outcome will be very precise and very accurate like that ok. So that is the overall idea. Now all this variability when I am talking about variability in the CTQs like that then I need to visualize the variations like that. I need to visualize the variation so that I can do something on that. So visualization is very important. So our overall objective is to reduce variation. I told that quality objective is to reduce variation assuming that it is on the target. So I want to reduce the variation. So variation reduction there are different ways of doing that. One of the I told that we we are talking about statistical process control which helps into eliminating the special cause or abnormal causes like that or assignable causes like that and design of experiment can also reduce the variations like that. Common cause variations can also go down when I am using design of experiments. And and we have one phase at the initial starting point of quality we have a acceptance sampling that means we are doing inspections we are trying to eliminate good pads and separate good pads good good good products with the bad products like that. So in this case what happens is that we try to segregate that one and we have certain plans acceptance sampling plan which is used which is also using statistical techniques. So over here and which gives you a a basis of selecting an item and not selecting an item from the vendor or suppliers like that ok. So this is only post modeling products have been produced by the vendor and it has come to your end and I want to reduce the variation over here. So I I tell them that you segregate this is not good and so in that case you give me good products like that I have a acceptable quality label. So you have to achieve that one then only I will accept your product. There can be small rejections like that but there is always a probability of rejection but we will if you reach this this quality label and then only we will we will do sampling inspection. Inspections can be 100 percent inspection also totally segregate good and bad like that but because of economic constraints sometimes what we do is that and also suppliers improves like that. So we go for sampling inspections like that. So in that case not all complete products will be taken some part of that will be taken and based on that I will make a decision whether to accept but I am taking a risk over there ok. So one is to reduce the variability in the input conditions what we can do is that we can implement inspection over here. So inspection is one of the quality aspects which generally the implements inequality program. So we start a quality initiative. So initially you may you may do inspections like that. Then you implement statistical process control to remove SNA will cause then design of experiments to do further improvements like that. So that way we are reducing variability like that ok. So but remember that inspection is a post-mortem activity nothing is done on the process so only thing is segregation is happening over here but other things are happening in the process statistical process control design of experiments all are happening in the process ok. So now what that is diagrammatically illustrated over here what you see is that variability of the process this is like a normal distribution you have heard of this in statistical codes like that which is having a mean of mu over here this population mean over here and with a standard deviation sigma like that. So this is just population information over here what you see diagrammatically and there will be specification limits over here upper and lower limits what is specified over here what will happen in inspection there will be rejection because I am not doing 100 percent over here. So there will be products which are bad like that so there will be. But some amount of variation can be reduced over here so overall variation will be high if I am doing only inspection over here sampling inspections like that. So I cannot do much on these variations over here but when I am using statistical process control like that applications of this type of techniques over here what I can do is that I can eliminate special cause variability like supplier to supplier there is huge variation over here I can remove a supplier which is a giving a high variability products like that so I can remove that one that is the assignable cause or special cause what we can eliminate out of the process ok. So that way we can identify which is creating more problems they uses a visualize to visualization tool in statistical process control and that will indicate which is normal which is abnormal scenarios and those abnormal scenarios we take action so that it does not recur. So that is known as that tool is known as statistical visually we can we can differentiate between normal scenario and abnormal scenarios like that that is known as statistical process control technique that we use over here. And the third one what we do is that we have a design of experiment that I want to reduce it further. So in that case what is required is systematic experimentation or statistical experimentation that is known as design of experiment DOE we can we can write over here as design of experiments over here that is the final phase of quality. So we we do design of experiments over here ok. So our objective is to understand so you see that the variation is represented like normal distribution but we have shown also histogram as the variation. I want to see where is the where is the mean where how much is the standard deviation like that. So visualization and also try to figure out the what is the value of mean what is the value of standard deviation like that. So for that visualization of the data is important so what we will do is that we will try to use visual tools try to understand how to visualize using a specific software which is over here Minitab but there are other softwares where we can visualize like that R interface can be used like that and you can use SAS or any other interface but here we will talk about how do you use Minitab in quality control and improvements like that ok. So first we will try to do basic things visualization of the data that means we will try to draw a histogram in Minitab that will give you some sense and then we will go ahead. So for this session we will we will discuss about how to implement how to visualize using Minitab some data set which is which we want to which can be a CTQs like that so ok. So histogram what we have what we have told is that it will have frequencies on one axis over here and it will have CTQs defined on this axis x axis will be CTQs and y axis will be frequency observations over here. So there will be this will be bins like that and in this case class number of class will be there 1, 2, 3, 4 like this n number of classes will be there and interval will be there in a class there will be intervals like that and there were highest frequency will be plotted over here. So this will be like and if you join the midpoints over here you can you can see some some belt shape kind of curve that you can see over here which may be in normal distributions like that. So if you join the midpoints of the histogram each of the bins like these midpoints you can you can you can just join them and you can see what is the shape of the or shape of the data set basically of the data set like that ok. So in this case frequency on one axis and one axis is CTQs like that. So let us try to plot this one in Minitab. So I will just introduce the Minitab software to you and then in next session we may go ahead with using that ok. So when this Minitab software can be downloaded from websites like that so I am showing you the websites maybe from Minitab used you can start a 30 days free trial like that and try to download the Minitab software and 19 version is available I think. So this Minitab 19 version there can be earlier versions also somebody may have installed earlier versions like that and maybe 17, 15 like that. So it is they continuously update the recent one like that 17, 18, 19. So I will use the 19 one more or less that is the latest one what I can understand at present. So that that we are for academic purpose we are using this Minitab 19 for illustration over here ok. So whenever you install Minitab from that website you download that one and you install that one what will happen you get a icon over here in the desktop like that. So if you double click that icon what will happen is that interface will open and in that interface we will be working using that interface to visualize the data over here ok. So there are three areas over here what you can see is that this is one of the area where the data set can be can be you can write down the data sets over here or name the data set as X over here and you can write the numbers over here. So see this row one here from here the data will start basically it is the column headings that you can provide over here X and Y like that we can provide headings over here you can write the numbers over here whatever number you think out and you can also paste it from excel sheets like that. So if we take it from excel let us say some Minitab files and you can open some excel files maybe a quality course marks we want to paste that one in this. So what we will do is that we can we can just this is in excel sheets what we are I can copy this column over here control C and then what I can do is that I can I can paste the data sets over here and control V over here always remember that the first first row over here is the title of the column like that. So you cannot paste number over here. So number starts from row number 1 like this 1 2 whenever 1 here from you can write the numbers like that you cannot paste numbers over here and Minitab cannot analyze that one. So in this case so these are the here you can place the data like this. So in this case I have placed in C 3 columns like that and then I can save this file also that means I can save this this is known as worksheet in Minitab this worksheet can be saved. So if you go to file and then you go to save worksheet as and let us say I am I am saving it in desktop and I am naming it some test files like that. So whenever you do that you save that one so this will be saved as test MVX what you see on the on the below over here and this files can be called time and again to do the analysis like that. So and if you close this one let us say I am closing this one and they it will ask projects can also be saved. So we will talk about projects but I am not interested in projects at this time point because I want the data set only to analyze because analysis can be done multiple times like that. So if I get the data that is sufficient for me to analyze like that. So I am saying no to the project savings like that. So I go out of this and then what I can do is that later on I want to see the data set and again analyze this one. So I open so this is saved in desktop and with a file name test over here and I have opened that file by double clicking that file over here. You can see that the marks are stored over here. So if you change the marks over here let us say 66 and then I want to save this one what you can do is that go to file save worksheet as and then just replace this file over here test and you replace the file it will ask whether you want to replace that one. You replace that one and immediately what you see is that it is replaced it has become 66 now and now you can analyze analyze the data like that. So if you have to do the analysis on the top you will find options over here statistics and there are various options that you can see over here like graphs you have many options one of the options is histogram over here. So whenever you click histogram then we can draw histogram of this data set like that and whenever you click histogram over here and then analyze and then click ok like that and what will happen is that that we will take in the next session what will happen this session window here you will get the data sets like that. So I am just showing you just for illustration over here more details we will do it afterwards. So I am clicking histogram over here from the top down menus over here I want the graph histogram of this data set of quality marks. So I will go to histogram and I will say simple histogram draw the histogram of quality codes like that. So then it will ask which variables over here. So I will double click this variable my cursor should be over here in graph variables. So it will blink in this graph variables then what I will do I will hide double click this one C3 and I do not I am keeping all as default in mini tabs like that later on we will see what are the options that we have then I click ok over here immediately this in the session window that you see this is the session window what you see. So this is the session window of this projects over here. So this is the distribution of marks or histogram mini tab has drawn automatically. So if you double click this one in the graph so immediately it will enlarge this one and you can enlarge this one you can copy the graphs also. So you can you can edit the graph also. So if you want to edit just click double click this one and if you want to change this histogram of marks you can do that and click ok. So many possibilities I can change frequency if you want to see in one column what is the data values within that bins like that. So the bin is starting from if you if you click this one the bin is starting from over here 62 this is coming out to be this is coming out to be this bin will come out to be. So in this case let us let us enlarge this one so that we can. So this bin is giving you 67.5 to 72.5 like that. Mini tab does it automatically histogram there are rules to define class and class intervals like that number of class and class interval based on the data set. So those calculations can be seen in books. So in this case what I am interested you know where is the central tendency of the data around 70 we can see the central tendency of mean of the data set. I need to understand the varying variants of these data so that we will talk about how to see that one. So only thing is that what I am saying is that you can draw the the diagrams will be plotted in the session window then you can copy this graph like copy if I want to copy this one and paste it in some other interface. So if you want to I have copied this one I want to paste it in excel like that. So I will right click this one and I will paste this one so automatically mini tab this diagram has come in the excel like that ok. So in reports any reports you want to place what file or any other files you can just paste it pen shop anywhere you can take this one and you can just paste this diagram like that ok. So that is all what we want to discuss in this session for this session 4. So we will continue with visualization of the data so we are talking about accuracy and precision for that we are using histogram one thing we should remember that this is for continuous data that we are doing over here. So in this case what we are doing is that we are assuming so in this case what we are doing is that we are assuming this data set what we have CTQs that we have written over here. So this CTQ is basically is a continuous variable. So this is continuous so that means within the scale it has infinite possibility number has infinite possibility only the least count of the instrument that is measuring this strength and that data I am I am just monitoring like that. So in this case maybe second place of decimal or one place of decimal like that so it depends on risk count but there are infinite possibilities of measurement within a range of let us say 150 to 170 infinite possibilities of numbers can be measured. So this is a continuous data and that is why I am drawing a histogram. Histogram is for continuous data whenever I have a continuous data I can use histogram to see the mean and variation of the data. How is the spread of the data what is the shape of the data like that where is the central tendency like that all these things can be seen visualized like that numbers can be calculated numbers can be seen by some other options like that and spread also can be seen by some other options like box plot what we will discuss in the next session like that. So we are discussing our visualization of the data. So we will continue because this data visualization is very important in quality if I can visualize then I can I can take some corrective actions based on the data and then I can interpret something like that. So this is known as we can say descriptive statistics of visualization of the data and based on that we we make some initial conclusion about the data or make some initial assessment of the data basically ok. So we will continue from here in session 5. Thank you for listening so we will continue with data visualization in quality. Thank you.