Hi, guys. Welcome back. I'm Katherine McKeever, and this is your lean six Sigma green belt. Today we're gonna go over hypothesis development.
So we have completed our root cause Analysis activities and Weah's greenbelts have an idea of where our starting point for our hypothesis is going to be. We talked quite a bit about hypothesis diddle amending yellow belt, so I would recommend you go back and take a peek if any of this seems familiar.
But you was a greenbelt, a practitioner or greenbelt facilitator. I want you to be able to understand how to write hypothesis statements,
and I want you to be crystal clear on the difference between your null and your alternate hypothesis
so in yellow. But we said What is a hypothesis? And we're like, Well, it's a question. It's a starting point. It's a relationship. Now you're in greenbelt, you're more sophisticated, and you know that your hypothesis is asking whether or not the root causes that you identified are in fact,
independent variables on your outputs or your problem statement.
So now you're using your hypothesis development to take your root causes that you identify and say, Do these things really change our output. Do we get what we're looking for in our dependent variables or process outputs?
This is where we're going to start using our data to explain things.
So it is a proposed explanation until we test it. But something to keep in mind we talked a little bit about in Yellow Belt, and I will give you a tool that will help with us in a few lessons. This does not account for correlation and causation. So if you remember we said, Correlation says, Yep, there's a relationship,
but it might be with 1/3 variable
that we're not seeing.
And causation is what we're looking for as lean six Sigma practitioners where it says, Yep, this F of X does, in fact equal. Why so correlation? Remember ice cream sales and shark attacks all happened during the summertime
causation. What we're looking for will have a tool to help tease that out
in a couple of lessons.
All right, now you're agreeing about your writing them. You did a really good job because you listen to my advice and you documented all of the potential root causes during your root cause analysis, and you've got a sense of what do you think is going to be the one you should look at first? Let's start writing them.
So the first important thing that you need to know is
they must be related to your project. Objectives I get that feels like it should go without saying, but just trust may they must be related to your project objectives. So what that means is, if your project objective is throughput or how many things go through the process,
preferably in a set amount of time,
then your hypothesis needs to be focused on throughput. If your objectivist cycle time, then your hypothesis needs to be focused on cycle time, how long does the process take to complete? So
for the sake of this lesson, we're going to work with the hypothesis about removing steps eight through 11 for our process throughput.
Eso we've We know that we're drawing from a process map here as our root cause analysis we've identified these steps are either waste or non value add, and we say we're going to remove them, so our hypothesis is removing steps eight through 11 will increase process throughput.
Please note the phrase
The reason Why would This is important to call out is because, well, increasing process throughput is what we want as their project objective.
But phrasing matters.
So when we are looking at are null and our alternate hypothesis How we choose our words is very important. So you're no hypothesis is what happens if nothing changes. This is the status quo. This says all of this hard work you did didn't matter.
So the phrasing that you're going to use is not
increase. So you've taken a stance and we've said removing step eight through 11 gonna do something. This is what we're going to do. So you're no hypothesis says removing steps eight through 11 will not increase throughput, not decrease it not, we're going to say stay the same.
is because of the fact the co founding factors, those other things that may cause correlation but not causation. So what? We were leaving this option open that there is more out there so much like we're really good about the phrase explicitly saying what isn't gonna happen.
We're also good at saying that we failed to reject the no hypothesis or we rejected the null hypothesis. This is very important
because we're not saying that our alternate hypothesis worked were saying that under these conditions, removing steps eight through 11 we were in fact able to see a difference from the no hypothesis.
So when we reject the null hypothesis,
we say our alternate hypothesis did something. When we fail to reject the null hypothesis, we say the null hypothesis is still here. But there is a chance that something else will change it.
Until we reject the null hypothesis. This gives us an opportunity toe work through our solutions. So you will, more than likely, as a practitioner, be testing a lot of different hypotheses. So removing step one steps eight through 11 removing steps 11 through 13.
You know, if we re program our computer, will it see this?
Those sorts of things So important things from this slide no hypothesis, status quo alternate what we think will happen. This is our okay. If we do this and we implement our great solution, we're gonna see this wording is important.
So increase throughput. This is going to be more is better if you have let decrease, that's going to be less is better if you have nominal is best you want to eliminate. So if you remember back when we were talking about understanding our metrics, this one
So we're going to want to increase our throughput, which means that we're processing MAWR units through a set amount of time in our process. So we're able to complete Maurin the same amount of time.
Wording is so important because because we go back to our distributions. We talked I talked a little bit about We want our distributions on a continuous spectrum because our statistics 68% 95 98 This is going to give us our probability
that we're going to be able to complete our process to our customers specifications.
CBK PPK Um, the qualifiers in your hypothesis statement will tell you what side of the distribution or what behavior you are looking to see in your distribution.
So if you are looking for an increase, you are looking for a positive shift where your numbers get larger.
Remember, we talked about positive and negative Skewed nous has to do with longer tails. You want to move your longer tail to a larger number, so increasing throughput time. We want to increase the number of widgets through us that time. We want to see a positive shift.
If you want to see where nominal is best or less is better,
you want to see a negative shift. You will decrease of the throughput time. And let's say that you're an indecisive type of person or a management researcher, and you don't know what's gonna happen. But you think something's gonna happen. Then you're going to say that there is a relationship
between steps eight through 11 and throughput time or that, UM,
removing steps eight through 11 will change throughput time that opens up both sides of the distribution for study. I tend to be very parsimonious or efficient with my time, So I like to say, Take a stand. It's going to increase and Onley tested the increase because the decrease
week, I mean, we kind of want to know, because I means that we made it worse.
But we know if it doesn't get better, we're not going to implement the solution to take a stand and just test that. But
traditional researchers will say test both sides first. If you see a change, then go ahead and retest with one side. It's however, you like to spend your time.
So with that, today we went over hypothesis development. You know the difference between an alternate and a no hypothesis, no means nothing's happening. You understand the significance of the warding and your phrasing because you now know that your look but what you're looking for in your project objectives
are what you want to look for in your hypothesis.
And you know that the way that you ward it will reflect in your distribution. This is how you know if your solution works when you're distribution, behaves the way your hypothesis says it will. So with that, we're going to move into our next module, which is testing. So how do we know that this hypothesis is, in fact
rejected or failed to be rejected?
So I will see you guys there