Atypical Distributions

Video Activity
Join over 3 million cybersecurity professionals advancing their career
Sign up with
Required fields are marked with an *

Already have an account? Sign In »

9 hours 53 minutes
Video Transcription
Welcome back, guys. Uncapher MacGyver in This is your lean six sigma green belt. Today we're going to go over a typical distribution.
And I love a typical distributions because this is where the puzzles are. This is where the analysis is. We look at our normal distribution and we're like, Yep. If we do, if we pull this lever or we change this X will see why
this is great. This is very helpful. This is the majority of your organization's work.
Hopefully. But a typical distributions is where we're gonna use a lot of our root cause analysis because it's not functioning the way we're expecting.
So we're going to go over a lot of them. Eso you When you finish this, you'll be able to recognize the common atypical distributions so you'll see skew. You'll look at by motile We'll look at hockey stick Well, look at you shape and then we'll recognize uniform distributions.
So by the end of this, you have a good sense of what normal looks like. You should have a good sense of what abnormal looks like
so skewed. We've talked about this before. We talked about this when we looked over distributions because skewed nous is the most common. A typical distribution, and it's where you're going to do the majority of your root cause. Work is agreeing about, like harsh but true skewed distributions are going to be what pays your bills?
Um, it we know that it is skewed when mean does not equal median.
We remember negative and positive from our distribution has to do with where the tales are on the dog. So where is the longer side of it? Can be negative or positive. We know this so skewed. What Skewed tells us is the majority of our data is performing on one
or another side of the average.
That means that if we're thinking about how our data is performing, we are seeing outliers where we are not seen normal, where 68% falls within that one standard deviation of the mean, which tells us that there is something affecting our process.
We'll talk about trends a little bit when we get further on, when we're looking at our data points in our trends, statistical process control,
but skewed simply told us that the majority of our data is clustering on one side versus the other. And we don't want this
unless we dio eso. If you are working in a type of organization or a type of process where nominal is best, you are going to want to see a very strong positive skew where your outliers pull your tail. But the majority of your data is closer to zero
so that, in that case, this is a normal distribution
if you wanna work. If you want to look at the distribution where higher is best, then you're gonna want to see a negative skew. Where you have a few that are below are mean, but the majority of the case is above. That's what we want.
So with that takes skewed nous with a little bit of a grain of salt. And think about what you're
process of or your project objectives or your process Goals are. If you need a refresher, go back and check on your charter. You can see what was the problem statement. And what are the goals that will give you a sense of the Is this normal for this process, or is this what we want from this process?
This graph are by motile distribution. We never want. And I love seeing this. This is so cool as a black votes or a monster by felt mentor. Because what this tells us is we have two data sets. So I love this because this is a great learning opportunity
in our data collection system.
So when we say indicate two data sets, remember that lean, started or lean six sigma started in manufacturing. So the example that we always teach is two different lines producing the same product. You can look at different shifts in our case because we tend to be more knowledge workers. At this point,
you're going to be looking at different teams. So remember when we talked about our segmented data where we take equal numbers from different teams? Let's say that's the logical segment from that, but we see this. What we're seeing is completely different processes. This means that from, um, capabilities maturity model back in yellow Belt,
that groups of people that do the same work are not doing it the same way.
So we're not quite at our threes and fours warm or in are ones ensues. So this is a very diagnostic distribution. When we see it.
It can be any shape so flat as a pancake. Super toll. What you're gonna look for in this is two distinct peaks. So we're going. What that indicates to us is we have two unique sets of descriptive statistics specifically to unique averages, which will then give us
the dispersion of accordingly. And then you'll see them lumped together
where the data overlaps
hockey stick distribution. This one's also called RJ's shaped distribution. So another distribution where we have to be really thoughtful about, what does this tell us? It indicates runaway growth. So if we're talking about revenue, we want this.
This is the coolest thing ever. This is what venture capitalists look for
Is hockey stick distributions on our revenue? However, if we're looking at, say mistakes, we do not want this because this means we were doing OK and then boom, our mistakes shot up. So then we're gonna want to look for a cause going back to our root cause analysis.
So something to be mindful remembered you, as a practitioner need to use your discretion when you're looking at these insane.
Is this really abnormal, or is this what we want? Hockey stick is a great way where I have seen this Really common is when systems go down. So if you look at like, customer complaints,
doing okay doing okay, Boom systems down. Now everybody has something hateful to say. And again, of course, venture capitalists. Look for this. When we look at revenue projections.
Next one up, you shipped distribution. What this one indicates is that you, your extreme values air higher. This is the opposite of a normal distribution where the majority of your data hangs out in the middle.
The majority of your data in this case hangs out on the end. It is shaped like a rounded, normal distribution. If you flip it upside down. Where I have seen this in reality
is in hospital costs. So if you think about, if you're going to hang out at a hospital, hopefully you're not sick or if you are, it's very easy to resolve. But if you go to a hospital, you have ah, high level of work at admittance. So we have all of these nurses doing your admittance. We order tests, we do a flurry of testing.
And then
there is that calm. Your tests are being reviewed your hanging out taking in the bad daytime TV. There's not a lot of cost associated with it. Will you have your hospitals operating costs? We know that nurses are always going to be there, but it's not going to be a great strain on the resources. And then
it comes time to discharge.
And there's another flurry of work. We've got to get your discharge instructions. We've got to get your prescription. The doc has to come in and tell you what happened on your results. So then you have a higher number of hospital resource is being associated with a patient. So that's where I've seen the U shaped distribution is more breaking down. Cost during hospital stay
the uniform distribution.
I have never seen this in reality. The only place that I have ever heard of this in reality is in planned communities. But what unit a uniform distribution tells us
is is that each value has an equal probability. So where it shows up in planned communities is if you think about houses with one bedroom and houses with two bedrooms and houses with three bedrooms.
Planned communities usually say 1/3 gonna be this four, Plan 1/3. This floor plan 1/3 of this floor plan Last 1%. Our show room. Um, I again, I have never seen this in reality. If you do see this,
please paying me because I want to see it too.
Um, not very common, but it is something to note. My advice would be if you saw this. We should probably re measure because there's probably something funky. Either we're not working with the right data type or there's something in our measurement system or sampling.
So with that, if you see one of these and it is not good, so it's not hockey stick revenue. The first thing that you need to do is repeat your data collection. So you want to go back and you want to make sure it is special cause versus common cause
so common causes kind of stuff that happens every day. Special cause isn't something that specific that happened. That's what we're looking for. Both cause types
when we do root cause analysis, but they give it we have different interventions on, and then the next thing that you want to do is perform a root cause analysis. So If you see something that doesn't work the way that it should, like a hockey stick for customer complaints, you're gonna want to do a root cause analysis.
What exactly started
these customer complaints?
It's today. We went over a typical distributions. We know that that means that they're not normal distributions and they can be negative. But they can also be a good thing.
So we need to recognize as practitioners. How does this distribution tell a story compared to our project objectives or process goals? And then we know how to recognize. Skewed by motile hockey stick U shaped and uniform distributions basically the whole enchilada. That is not the normal distribution.
But since we've looked at that, we're gonna shift gears and we're going to start talking about root cause analysis
for the next few modules.
Up Next