One sided F-test in Minitab

[tahirawan11]

Hi,

Can anyone tell me the procedure to perform a one side F-test in Minitab v14. The reason to perform this test is that i have two processes 'new' and 'old' and i want to test statistically that the 'new' process is less variable than the 'old' one.

/tahir

[harry]

Thread bumped for further response.

[Miner]

Minitab offers several options. Bartlett's Test, the F-test and the Levene test for equal variances. The first two assume and are sensitive to normality in the data. The Levene test is more robust to non-normal data.

See Stat > Basic Statistics > 2 Variances or Stat > ANOVA > Test for Equal Variances and Minitab Answer ID 1183

Note: The is no such thing as a one-sided F-test. These tests verify the assumption that the variances are equal. There is no assumption that one is larger than the other.

[tahirawan11]

Hi Miner,

Thanks for your reply, but as per NIST website, see (http://www.itl.nist.gov/div898/handb...on3/eda359.htm),

"This (F-test) test can be a two-tailed test or a one-tailed test. The two-tailed version tests against the alternative that the standard deviations are not equal. The one-tailed version only tests in one direction, that is the standard deviation from the first population is either greater than or less than (but not both) the second population standard deviation . The choice is determined by the problem."

so i am interested to know if Minitab offers a way to do a one-tailed version?

br,

/tahir

[Dave Strouse]

tahirawan11,

Looks like MINITAB doesn't do this. Probably because you don't need to.
Why do you want to do a one tail test?

From the graphics in the program, it is obvious which variance is larger.

BTW, if the F and Levines differ in signicance, I would rely on Levine's unless you have tested normality of the data.

[Miner]

Originally Posted by tahirawan11

Hi Miner,

Thanks for your reply, but as per NIST website, see (http://www.itl.nist.gov/div898/handb...on3/eda359.htm),

"This (F-test) test can be a two-tailed test or a one-tailed test. The two-tailed version tests against the alternative that the standard deviations are not equal. The one-tailed version only tests in one direction, that is the standard deviation from the first population is either greater than or less than (but not both) the second population standard deviation . The choice is determined by the problem."

so i am interested to know if Minitab offers a way to do a one-tailed version?

Interesting. I learned something new. As you seem to be aware, Minitab does the two-tailed test. It does not have an option for the one-tailed version. The calculation of the F-statistic itself is the same for both the one and the two tailed tests, so you can use Minitab's calculation of the F-statistic.

You will have to ignore the p value and manually determine the critical F-value from tables then perform the comparison shown in the NIST Handbook.