Locked lesson.
About this lesson
One sample tests are tests of a single dataset that compares the descriptive statistics of that data set against target values.
Exercise files
Download this lesson’s related exercise files.
One Sample Test Exercise.xlsx10.9 KB One Sample Tests Exercise Solution.docx
544.7 KB
Quick reference
One Sample Tests
One sample tests are tests of a single dataset that compares the descriptive statistics of that data set against target values.
When to use
One sample tests are used in the Measure and Analyse phases to determine if the current performance of a process is meeting expected values. One sample tests are used in the Improve and Control phases to determine if the changed process is meeting target values.
Instructions
One sample tests are easy to perform. Descriptive statistics are calculated for the sample dataset. These statistics are compared to the target values to determine if any differences are statistically significant. In Lean Six Sigma projects, one sample tests are used to check both the mean and the variance or standard deviation.
The form of the hypotheses for one-sample test are:
- One-sample T Test
- The Null hypothesis is: H0: x̄ = target mean
- The Alternate hypothesis is: Ha: x̄ ≠ target mean, in some cases the alternate hypothesis will specifically focus on greater than or less than the target mean.
- One-sample Variance Test
- The Null hypothesis is: H0: σ = target standard deviation; σ2 = target variance
- The Alternate hypothesis is: Ha: σ ≠ target standard deviation; σ2 ≠ target variation. In some cases, the alternate hypothesis will specifically focus on greater than or less than the target value.
One-sample T Tests compare the mean of the dataset to a specified target value. The data does need to be normal data. If the raw data is not normal, use the central limit theorem to normalize it.
-
- Excel cannot do this test.
- Minitab:
- Stat > Basic Stats > 1-sample T
- Enter the column containing the data
- Check the box for Hypothesis test
- Enter the target value
- Use the option button to use a “greater than” or “less than” alternate hypothesis
One sample variance test compares the variance or standard deviation to the specified target. The data does need to be normal. If the raw data is not normal, use the central limit theorem to normalize it.
-
- Excel cannot do this test.
- Minitab:
- Stat > Basic Stats > 1 Variance
- Select the data column
- Check the box for hypothesis test
- Select whether targeting the standard deviation or the variance
- Enter the target value
Select the Option button to use a “greater than” or “less than” alternate hypothesis
Hints & tips
- The one sample T test is somewhat forgiving if the data is not normal, as long as it is symmetric.
- You can fool Excel into doing a One Sample T Test by creating a second dataset of nearly all zeros (you need at least two values in the dataset) and then doing a 2-Sample T Test with Unequal Variances from the Data Analysis toolpak.
- Although the purist will prefer to use the variance instead of the standard deviation, the standard deviation is perfectly acceptable for use in this test.
- The one sample T-Test is good for checking for special cause factors that shifted the process output.
- The one sample variance test is good for checking for common cause factors that have impacted the variability in the process.
- 00:04 Hi, I'm Ray Sheen.
- 00:05 The next set of tests are used to understand to what degree you can use
- 00:10 a dataset, is it representative or not?
- 00:14 I'll start again with the hypothesis test decision tree.
- 00:18 We are working with normal, discrete, and continuous data.
- 00:22 These are the tests that are used when there is only one dataset for analysis.
- 00:27 Let's take a minute to talk about testing with only one dataset sample.
- 00:32 These tests are normally done to demonstrate that the dataset is part of
- 00:37 a larger population, or is not from that same population.
- 00:40 But the analysis team does not have access to the larger population data.
- 00:45 They only have access to some descriptive statistics that represent that data.
- 00:50 So to conduct one sample test,
- 00:52 we need to have the descriptive statistics from the larger data population.
- 00:56 That might be historic statistics, such as the demonstrated process
- 01:01 performance when the new system was first installed.
- 01:04 It could be tied to industry comparison data.
- 01:07 For instance, if you're working in a call center, you would want to compare
- 01:11 your process performance to the industry average data for other call centers.
- 01:15 Sometimes it's not even based upon an actual population, but rather the data is
- 01:20 being compared against a target or goal that was embedded in the project charter.
- 01:26 Once you have the target value, now you can test to see if the mean of your
- 01:30 dataset matches the target mean or if the variance matches the target variance.
- 01:36 Let's think about this in the historical context,
- 01:38 since that's the way I've traditionally used this test.
- 01:42 You could check your mean against the historic mean to see if there's been
- 01:46 a shift in the process performance, either better or worst.
- 01:49 A shift in that performance would imply a special cause.
- 01:53 Or you could test your variance against the historic target variance to determine
- 01:58 if your standard deviation was smaller or larger.
- 02:01 A change in variance implies a change in the common cause factors that
- 02:06 are part of the routine process performance.
- 02:09 If you start with the one sample T test,
- 02:12 the test is somewhat similar to the one sample test of proportions.
- 02:16 Only instead of a target proportion, the target is a value for the mean.
- 02:21 The one sample T test checks the sample to see
- 02:24 if the mean is statistically different from the target value.
- 02:28 This is often used to test to see if there's a real difference in a product or
- 02:32 process after an improvement has been incorporated.
- 02:35 The target value being the original performance before the improvement.
- 02:39 Of course, we're working with normal data.
- 02:42 Otherwise, the mean value is not a meaningful number.
- 02:45 The hypothesis statements are very straightforward.
- 02:48 The null hypothesis is that the sample mean is equal to the target mean,
- 02:52 there's nothing unusual or special in the sample.
- 02:55 The alternative hypothesis is that the sample mean is not equal.
- 02:59 It could be specifically stated that it is greater than or less than the target.
- 03:04 The T statistic is a simple calculation,
- 03:07 the difference between the target and the actual value.
- 03:10 This is divided by the standard deviation over the square root of the number of
- 03:15 items in the sample.
- 03:16 So let's take a look at the calculation.
- 03:19 Excel does not provide a one sample T-test function.
- 03:22 So if you're working with Excel, calculate the T statistic using the formula from
- 03:28 the previous slide, then compare the value to the t value in the table shown here.
- 03:33 Enter the table using the alpha value and the number of degrees of freedom.
- 03:38 Minitab has a function, start with the Stat pulldown menu,
- 03:42 then select Basic Statistics, and select 1- Sample T, and
- 03:46 enter the column with your sample data.
- 03:49 Then check the box for hypothesis test,
- 03:52 and enter the target value in the field labeled Hypothesis mean.
- 03:56 The default is to check if the mean is not equal.
- 03:59 If you want to specifically check for greater than or less than, then select
- 04:04 the option button and change the field, just as we have done with other tests.
- 04:09 Next, we will consider the one sample variance test.
- 04:13 This test is very similar to the one sample T test only checks the sample to
- 04:18 see if the variance is statistically different from the target value.
- 04:22 This is often used to test if there's a change in the process environment and
- 04:26 process management.
- 04:28 It could be to see if there were any degradation over time, or
- 04:32 it could be used in the improve phase to see if the improvements reduce the common
- 04:37 cause variation, the target value being the original performance.
- 04:41 Of course, we're working with normal data, otherwise,
- 04:44 standard deviation would not be a useful measure in this test.
- 04:48 The hypothesis statement could use variance or standard deviation,
- 04:52 I've seen it both ways, although the purists would say that variance is better.
- 04:57 The null hypothesis is that the sample variance is equal to the target variance,
- 05:01 nothing has changed to affect the process variability.
- 05:05 The alternative hypothesis is that the sample variance is not equal.
- 05:09 It could be specifically stated that it is greater than or less than the target.
- 05:14 Excel does not provide a 1-Sample variance test function.
- 05:18 If working with Excel, use the formula for Chi-Squared variance.
- 05:22 This is the sample deviation squared,
- 05:24 multiplied by the number of items in the sample -1.
- 05:28 That is then divided by the variance of the target population.
- 05:32 Minitab has a 1-Sample variance test.
- 05:35 Start with the Stat pull down menu, then select Basic statistics, and
- 05:40 select 1-Variance.
- 05:41 Enter the column with your sample data, then check the box for hypothesis test.
- 05:46 You then select to either provide a target variance or a target standard deviation.
- 05:51 The default is to check to see if they are equal.
- 05:54 If you want to check for greater than or less than you must select the option
- 05:58 button and then change the field, just as we've done with the other tests.
- 06:03 Whether you're checking the mean or variance on just one sample of data,
- 06:08 these tests can tell you if something is different.
- 06:11 Either different from a historical perspective or
- 06:14 hopefully different in a better way after you've implemented your improvement.
Lesson notes are only available for subscribers.
PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.