I'm taking stats with calculus at university right now and this is extremely helpful! It's concise while still providing all the necessary information. Perfect for using in conjunction with homework or to review for tests. I'll definitely be sharing this with my classmates. Thank you!
Hey Ahsan, whenever you're dealing with a situation where you're attempting to determine if your parameter of interest "has not changed", then the best approach is to use a two-tail test. This creates a rejection region in the upper and lower tails of the distribution. This is the correct approach because your parameter could go up, or down, and you'd want to be able to reject that null hypothesis in either of those situations.
Hypothetical Question: Let's say I have 1 year worth of data indicating lead times for every unit produced. Project goal is to reduce average lead time. Several factors have been determined as potential root causes and I'd like to validate, whether or not they affect avg lead time. Null hypothesis is the Factor A does not significantly influence avg lead time. Alternative hypothesis is that it does. a) For my population distribution, can I plot the lead time frequencies over the full year of data, or do I need to generate random samples (let's say 30 samples) from this data and plot the frequency of their respective avg lead times (hopefully forming a normal distribution), before beginning my hypothesis testing? b) What if it doesn't form a normal distribution? c) Assuming a normal distribution, do I sample the avg lead time after we've piloted Factor A and convert that to a z-score to see if it falls within/outside our rejection regions? As a Lean Six Sigma Practitioner, I find your videos supremely valuable. Thank you for everything that you do!
Hey Samuel! Okay, so in terms of your population distribution you can use the entire data set. Then I assume you'll be taking your data (after the change) and comparing it to your population. You can run that hypothesis test without worrying about the normality of your population distribution, because the central limit theorem tells us that the distribution of sample means will be normally distributed. And it is the distribution of sample means that we will use to make the accept/reject decision for your hypothesis test. The other comment i'll make here, is that you have a population of data before the change, and I presume you'll have the entire population of data after the change. So instead of performing a hypothesis test, you can simply talk about the differences before and after the change. For example, you might see a 10% reduction in lead time, that might not trigger the "reject" conclusion in a hypothesis test, it still might be a 10% reduction in lead time. Does that make sense?
Hello, i liked this video but i wish you added more examples or problems to solve. The null hypothesis and alternative hypothesis are tricky. Didn't fully understand them but i got the main picture.
Hey There!! No, I don't think so, you should be able to calculate a z-score using a regular calculator. Try again to calculate the numerator first (top half of the equation), then calculating the denominator (bottom half of the equation) second, and dividing the first by the second.
