Hypothesis Test Process

Quick reference

• 00:04 Hi, I'm Ray Sheen.
• 00:06 We discussed the general concept that's known as the scientific method.
• 00:10 Well, now let's look more specifically at the hypothesis testing process.
• 00:15 The hypothesis testing process consists of five steps.
• 00:20 The first step is to define the problem or issue and the goal.
• 00:24 This is the same as stating the question in the scientific method.
• 00:27 The second step is to write the null and alternative hypothesis.
• 00:31 We will go through how to do that in the next lesson.
• 00:33 The third step is to collect the data for the analysis.
• 00:36 Often on a Lean Six Sigma project, this was already done in the measure phase.
• 00:41 So at this point it may be to just determine what data is available.
• 00:44 Occasionally, you'll have to go and collect more data.
• 00:47 The fourth step is to use the appropriate hypothesis test to calculate the test
• 00:51 statistic.
• 00:52 And the last step is to make a decision based upon the results of
• 00:55 the hypothesis test.
• 00:57 This decision is either to reject the null hypothesis or
• 01:00 fail to reject the null hypothesis.
• 01:02 Let me go through each of these steps in a little more detail to highlight some
• 01:06 aspects.
• 01:07 Step 1 was to create the goal of the hypothesis test analysis.
• 01:11 Specifically, our step was to define the problem or issue being investigated and
• 01:16 to state the goals.
• 01:17 On a Lean Six Sigma project, there could be many different reasons for
• 01:21 conducting a hypothesis test.
• 01:23 A common one is to determine if your data is normal.
• 01:25 If it's normal, you can apply SPC principles.
• 01:28 If not, you may need to transform the data before applying SPC.
• 01:32 Another common question is to test to determine whether a variable is really
• 01:36 an independent variable that influences the process response or
• 01:40 if that variable has no impact on the response.
• 01:42 And one of the most common reasons for conducting a hypothesis test is to
• 01:46 demonstrate that the proposed improvement really makes a difference.
• 01:50 The next step was writing hypotheses.
• 01:53 As we stated, it is writing a null hypothesis and an alternative hypothesis.
• 01:57 Each of these are statements that are essentially the opposite of each other.
• 02:02 Such as the mean value of data set 1 and 2 is the same, or
• 02:05 the mean value of data set 1 is smaller than that of data set 2.
• 02:08 The null hypothesis will always take the form of stating
• 02:12 that there is no statistically significant difference in the data based upon
• 02:17 the item being investigated.
• 02:19 This is essentially saying that whatever the difference is noted,
• 02:22 it's just based upon normal random variation in the process and
• 02:26 does not reflect anything meaningful with respect to the problem or question.
• 02:30 In contrast,
• 02:31 the alternative hypothesis is often the assumption we're trying to prove.
• 02:35 For instance, that the change that was implemented has reduced the average
• 02:39 process time from what it had been.
• 02:41 Both the null and alternative hypotheses should include these elements.
• 02:46 Clearly identify the population or data sets involved,
• 02:49 identify the dependent variable and all independent variables.
• 02:53 And finally, if appropriate, stating the direction of the relationship such as
• 02:57 greater than, less than, or equal to.
• 02:59 On to step 3.
• 03:01 This step is to collect the data for analysis.
• 03:04 Consider what data is needed to accept or reject the null hypothesis.
• 03:07 Does this data already exist in the data available to you?
• 03:11 If so, you're ready to move to the next step.
• 03:13 If not, you'll have some work to do.
• 03:15 Make sure that the sample data you have is from the population you stated in your
• 03:19 hypothesis.
• 03:20 Otherwise, you're not proving anything.
• 03:22 Of course if your hypothesis is that there are two different populations, you'll want
• 03:27 to make sure that you get data from each to show that they truly are different.
• 03:31 Many times the data was already collected during the measure phase of
• 03:35 your Lean Six Sigma project.
• 03:36 But to be sure it really does address your hypothesis, you may need to gather some
• 03:40 additional data to capture the effect of an additional independent variable.
• 03:44 Step 4 is the time for calculations.
• 03:47 In this step, we calculate the hypothesis test statistic.
• 03:51 Don't be fooled by the word test.
• 03:53 We're not in a laboratory conducting tests under a microscope.
• 03:57 These are tests of data sets.
• 03:59 It's all done with math and statistics.
• 04:01 Depending upon the hypothesis test, a different statistic may be calculated.
• 04:06 Some of the statistics are associated with the data set,
• 04:09 such as calculating the mean, the median, or the standard deviation and variance.
• 04:13 The analysis may be focused on calculating a correlation relationship or
• 04:17 determining if the data is normal.
• 04:19 And many of the tests calculate their own statistical value or coefficient.
• 04:24 These will be discussed as we look at the tests.
• 04:27 Our last step is to make the decision.
• 04:30 Based upon the evidence suggested we decide to either accept the null
• 04:34 hypothesis or reject the null hypothesis.
• 04:37 Notice we don't actually accept the alternative hypothesis.
• 04:40 If you wrote the hypotheses well,
• 04:42 rejecting the null will lead to accepting the alternative.
• 04:46 But a poor hypothesis could have a condition where both the null and
• 04:50 the alternative are not true.
• 04:52 So in our analysis, the actual decision is to reject the null or
• 04:57 to fail to reject the null hypothesis.
• 05:00 Let me give you an example.
• 05:02 The null hypothesis was that there were no differences between two different
• 05:05 data samples.
• 05:06 And of course, the alternative hypothesis is that the data sets are different.
• 05:10 The analysis of the data shows that the mean and
• 05:12 standard deviation is very different between the two data sets.
• 05:16 This is indicated by a very low P value which I'll explain in another lesson.
• 05:20 But the bottom line, we reject the null hypothesis that there is no difference.
• 05:25 And in this case, we can accept the alternative hypothesis that there is
• 05:29 a difference between the two data sets.
• 05:31 Following the hypothesis testing process will lead to a data-based and
• 05:37 statistically valid conclusion with respect to the initial question.

Lesson notes are only available for subscribers.