Retired course
This course has been retired and is no longer supported.
About this lesson
Graphical analysis techniques are particularly good for illustrating significance, similarities or differences, and correlation between parameters within a data set.
Exercise files
Download this lesson’s related exercise files.
Visual Analysis.docx97.4 KB Visual Analysis - Solution Part 1.docx
115 KB Visual Analysis - Solution Part 2.docx
91.8 KB
Quick reference
Visual Analysis
Graphical analysis techniques are particularly good for illustrating significance, similarities or differences, and correlation between parameters within a data set.
When to use
The graphical analysis techniques work very well when there is either a special cause problem or a singular root cause for the problem. In those cases, it is often faster to use the graphical visual analysis than a statistical analysis. Also, it is much easier to explain the problem to stakeholders and team members using visual analysis than with statistical analysis.
Instructions
Different types of graphs can be used for visual analysis depending upon the question or hypothesis that is being asked. Using a graphical analysis technique provides a visualization of the data that can quickly lead to conclusions about the problem.
When looking for significance of a parameter or significance within a parameter, use the vertical bar chart or horizontal bar chart. The Pareto chart, which is a special case of the vertical bar chart, is particularly useful. The pie chart can also be used in some cases, but it is normally used in families of pie charts, so either the charts will show that an item is significant in some instances (pies) and not in others, or that it is significant in many different instances (pies). Either of those may be very important information to prove or disprove a hypothesis.
Graphical techniques are also useful for visualizing similarities and differences. This can either be similarities between data sets using box plots or similarities and differences over time using run charts. In the first case, the box plots reveal which data sets are different, providing insight into underlying causes for the problem. In the second case, the run chart shows when a difference first occurred, providing insight into the occurrence of a special cause problem that must be addressed in order to get a stable process.
Graphical techniques can also show a relationship between two parameters. A scatter diagram shows a positive correlation if the data has a clear upward trend and a negative correlation if the data has a clear downward trend. If there is no trend in the data, the two factors are independent of each other. While the scatter diagram shows a relationship, it does not indicate which factor is dependent and which is independent. In fact, both factors could be dependent upon a third factor. Therefore, correlation does not mean causation, only that there is relationship.
Hints & tips
- Once the data is captured in a database or table, it is easy to create charts and graphs in both Excel and Minitab. It only takes a few mouse clicks, so visualize the data first before jumping into statistical analysis.
- Graphical or visual analysis is particularly well suited to single factor problems and special cause problems. That is why some of the basic problem solving methodologies rely solely on visual analysis.
- 00:04 Hello, I'm Ray Sheen.
- 00:06 Many times the fastest and easiest type of analysis that you can do
- 00:10 in a Lean Six Sigma project is a visual analysis.
- 00:15 I've already discussed many of the types of charts and
- 00:17 graphs that you can use to display data.
- 00:21 Let's now take a few moments to review them from the standpoint of data analysis.
- 00:26 Keep in mind that graphical displays of data are excellent ways to
- 00:29 communicate the data set to both stakeholders and team members.
- 00:33 I suggest that you introduce a graphical display of the data
- 00:36 early in a problem analysis session with the team.
- 00:39 Often the underlying cause will jump out from the picture of the data, and
- 00:43 the team knows what to focus on for further analysis.
- 00:47 The visual display does this by putting the different data elements
- 00:50 into the same frame of reference or context with each other.
- 00:53 Then, differences and similarities are easy to spot.
- 00:57 The picture draws the eye to the point of change or difference.
- 01:01 Or, if there is no significant difference,
- 01:03 the picture illustrates that the data is the same.
- 01:07 Now, different types of displays are better for different types of analysis.
- 01:12 And I'll spend the rest of this module going through these types of analysis and
- 01:15 the best displays for each.
- 01:18 Let's start with importance.
- 01:20 Some visual displays do a great job of illustrating what is important
- 01:23 within the data based upon the magnitude or size of what is shown.
- 01:27 This allows the team to quickly focus their analysis effort
- 01:31 on specific parts of the process or problem, and speeds the analysis.
- 01:35 The best charts for this are the Pareto chart, which is a special case of
- 01:39 the vertical bar chart, the horizontal bar chart, and sometimes the pie chart.
- 01:44 The team will also need to identify outliers or
- 01:46 strange data points within the data set.
- 01:49 These are often special cause occurrences and recognizing and
- 01:53 fixing those are normally the first step in improving the product of process.
- 01:57 So the team wants to find those as soon as possible.
- 02:00 The process will not be stable and
- 02:01 predictable until the special cause problems are eliminated.
- 02:05 Once those are identified and
- 02:06 the solution put in place, then outlier conditions can be removed from the data
- 02:10 set to allow a better understanding of the steady state process performance.
- 02:15 Charts that are particularly well suited for identifying outliers
- 02:19 are the run chart, which is a special case of the line chart, the box plots, and
- 02:23 scatter diagrams, all of which I will show on the next few slides.
- 02:27 In addition to importance, being able to recognize similarities and differences
- 02:31 across data sets can improve the quality and speed of your problem analysis.
- 02:37 When I find that two data sets are similar, I can combine the data sets and
- 02:41 often find significance within the combined data.
- 02:44 And the corollary is that if I see that the two data sets are not similar,
- 02:48 then the factor that separates these two data sets such as allocation, a person,
- 02:53 or a customer, or a supplier is a factor of significance and
- 02:56 requires further analysis.
- 02:58 In my opinion, the best visual tool for doing this analysis is the box plot.
- 03:04 Sometimes it's the occurrence of an event that
- 03:06 is the item of most significance in the analysis.
- 03:09 In that case,
- 03:10 you should use a chart that shows the performance of that parameter over time.
- 03:15 The point of change is often immediately apparent
- 03:18 as you can see that the data shifts from one level to another.
- 03:21 Or see the data start or stop a trend up or down at a certain time or
- 03:26 a certain unit number in the data stream.
- 03:28 These inflection points indicate something significant happened and
- 03:32 needs to be investigated.
- 03:34 The best chart for that is the run chart, as you can see in this example.
- 03:38 At unit number 9,
- 03:39 we see the data trend makes the radical shift in slope from downward to upward.
- 03:45 Another analytical conclusion we can quickly draw from data
- 03:49 is whether two parameters in the data are correlated.
- 03:52 In this case we can see if a relationship exists, and
- 03:55 if the relationship is a positive one or a negative one.
- 03:58 A positive or upward slope in the data shows that as one parameter increases,
- 04:03 the other parameter increases.
- 04:05 Whereas a negative or downward slope, indicates that as one parameter increases,
- 04:09 the other parameter decreases.
- 04:11 What this does not tell us is, which factor is the dependent factor?
- 04:16 Therefore, we still need to investigate this relationship,
- 04:18 to see if one of these is an independent variable that we can control in a way to
- 04:22 influence the other dependent variable.
- 04:25 Be careful, correlation does not mean causation.
- 04:28 It could be that both variables are dependent upon a third factor that has not
- 04:32 even been plotted yet.
- 04:34 Large trucks can pull more weight than small trucks but that doesn't mean that if
- 04:38 we just make the trucks bigger that they can pull more weight.
- 04:42 It turns out, they both attract size and
- 04:45 their pulling capacity depend upon the truck engine and transmission.
- 04:49 If we want to pull more, we need a bigger engine and transmission, and therefore,
- 04:53 we need a bigger truck to put around that engine and transmission.
- 04:57 But if I just look at the size and pulling capacity, and make a bigger truck
- 05:00 without a bigger engine, I might fail to get an improvement.
- 05:04 It can also be very enlightening to find that some factors are not correlated.
- 05:09 If we see that there is no relationship,
- 05:11 then we know not to put any further effort into that area and expect an improvement.
- 05:16 For instance, if we find that teacher certification does not affect
- 05:19 student outcome on standardized test scores, then we know not to waste time and
- 05:23 money investing in certification with the expectation that test scores will improve.
- 05:28 The most commonly used view for this type of analysis is the scatter diagram, but
- 05:33 it's limited to looking at just two factors at a time.
- 05:36 However, some advanced software can be used to plot more than two factors and
- 05:40 create a three dimensional surface plot.
- 05:44 Don't overlook the benefit of visual analysis.
- 05:47 Many of the basic problem solving methodologies rely solely
- 05:51 on visual charts.
- 05:52 They are especially good for finding special cause problems.
Lesson notes are only available for subscribers.
PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.