Subscribe Bookmark RSS Feed

JMP Script to do Hodges-Lehmann

bwwhite

Community Trekker

Joined:

Jun 18, 2012


Does anyone know of a JMP script to do the Hodges-Lehmann procedure on data? Certain European standards documents call for Hodges-Lehmann analysis to be done on specific kinds of data. The documents show how to do Hodges-Lehmann by hand. I can get the desired output using the 1-Sample Wilcoxon test in Minitab although it requires a lot of explanation about what parts of the Minitab output correspond to the desired H-L information. These documents specifically recommend StatXact or SAS to do Hodges-Lehmann. I don't have SAS, but do have JMP. I was hoping to find a JMP script to do Hodges-Lehmann. The test I'm performing is a one-sided test that the median of a data set is a specific value. And I must use Hodges-Lehmann.

Thank you!

Bruce

3 REPLIES
bwwhite

Community Trekker

Joined:

Jun 18, 2012

No one?

mewing

Community Trekker

Joined:

Jun 23, 2011

bwwhite,

Do you have a link to doing the H-L by hand?  Also, do you have R installed on the same computer as JMP?

bwwhite

Community Trekker

Joined:

Jun 18, 2012

The manual method that I found comes from a European standards document. Here is a link about the document. I'm unsure about copyright info, or I would post an image of the appropriate page or two. I was able to find a generalized algorithm here. I also found some SAS steps in this document and this one.

I believe I can cite a paragraph from the European document. After the median is found, the procedure below is described for a data set of size n=20.

"The mean pairwise differences that do not exceed the median ... are computed. From Table E.5 of critical values for Wilcoxon's matched-pairs signed-ranks test the entry for n=20 and a one-sided 0.025 level of significance the critical value of 52 is found. Hence c=52+1=53. The pairwise differences are sorted in descending order. The 53rd value is 0.11 [NOTE: this is from the sample data in the document]. hence the Hodges-Lehmann upper one-sided 97.5% confidence limit for the difference in lg Rs between RP and PP is 0.11, which is less than the agreed inferiority margin of 0.6. Therefore, the hypothesis of inferiority of PP is rejected and it can be concluded that the test preparation PP is non-inferior to RP."

The European document contains a small table of critical Wilcoxon values, but I have not been able to find a more complete table anywhere.

Thanks!

Bruce