turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Calculating Quantiles for a newbie

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Apr 9, 2009 8:27 PM
(3372 views)

To be honest I have never taken a statistics class and for the most part fail at math. I have a set of data that represents cycle time for several stages of an order and I was asked to find the weighted average which wasn't really a problem and I was able to do this inside of SQL. Then I wass told that it was great but they also wanted 80th and 95th percentiles for the orders. After looking confused for a few moments I asked them if they knew how to calculate it, what it meant, or if they had anyone else pull this before. The answer I received was "No but use JMP, we were told it could do it." Ok no problem, contacted IT found we have a license agreement, installed it, then I got to stare at it blankly for a few hours and trying to plug in a few numbers figured out that I have no idea what I am doing. I started looking for guides online but they all seem to assume a certain level of knowledge in mathmatics or JMP in general that I just do not posses. So hopefully someone can help.

The data I have looks similar to this:

Order OE-IP(x) Order IP - SHP(y) Order SHP - POD(z) Total orders(T)

5 ----------------------------- 2 -------------------------- 3 ------------------- 110

3 ----------------------------- 1 -------------------------- 7------------------- 2000

0 ----------------------------- 5 -------------------------- 4 ------------------- 200

There are about 800k lines of data like this so I won't post that up.

How would I use JMP to get the percentiles they are looking for?

For the weighted average I did

[(x+y+z)*T]=C for each line

(Sum of column C / Sum of column T) = Weighted Average

Any help would be greatly appreciated.

The data I have looks similar to this:

Order OE-IP(x) Order IP - SHP(y) Order SHP - POD(z) Total orders(T)

5 ----------------------------- 2 -------------------------- 3 ------------------- 110

3 ----------------------------- 1 -------------------------- 7------------------- 2000

0 ----------------------------- 5 -------------------------- 4 ------------------- 200

There are about 800k lines of data like this so I won't post that up.

How would I use JMP to get the percentiles they are looking for?

For the weighted average I did

[(x+y+z)*T]=C for each line

(Sum of column C / Sum of column T) = Weighted Average

Any help would be greatly appreciated.

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

In JMP the quantiles are specified in the Tables/Summary/ panel. You select the variable for which you want the quantiles and modify the quantiles percentage which is set by default to 25 and then choose Quantiles from the Statistics puldown menu. Grouping can also be set in the obvious manner.

--

David

--

David

4 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Looks like you don't need Jmp.

Here is how you calculate percentiles in SQL:

Suppose you have _Table_ that contains several groups with [Group_ID] of the values and you need to calculate 5th and 95th percentiles for these groups.

------ 5th percentile --------

SELECT MAX ()

FROM (SELECT top 5 percent, [Group_ID]

FROM _Table_

ORDER BY asc)

GROUP BY [Group_ID]

-------- 95th percentile -------

SELECT MAX ()

FROM (SELECT top 5 percent, [Group_ID]

FROM _Table_

ORDER BY desc)

GROUP BY [Group_ID]

If you still want to use Jmp, then go to Analyze ->"Fit Y by X', choose as "Y,response", [Group_ID] as "X,Factor", press OK. Then click on a red triangle in the popped-up window and select Quantiles.

This can be easily found in Jmp help, btw...

Thanks,

MikeK.

Here is how you calculate percentiles in SQL:

Suppose you have _Table_ that contains several groups with [Group_ID] of the values

------ 5th percentile --------

SELECT MAX (

FROM (SELECT top 5 percent

FROM _Table_

ORDER BY

GROUP BY [Group_ID]

-------- 95th percentile -------

SELECT MAX (

FROM (SELECT top 5 percent

FROM _Table_

ORDER BY

GROUP BY [Group_ID]

If you still want to use Jmp, then go to Analyze ->"Fit Y by X', choose

This can be easily found in Jmp help, btw...

Thanks,

MikeK.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Correction for the 95th quantile:

-------- 95th percentile -------

SELECT MIN ()

FROM (SELECT top 5 percent, [Group_ID]

FROM _Table_

ORDER BY desc)

GROUP BY [Group_ID]

Sorry,

MikeK.

-------- 95th percentile -------

SELECT MIN (

FROM (SELECT top 5 percent

FROM _Table_

ORDER BY

GROUP BY [Group_ID]

Sorry,

MikeK.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

--

David

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Tabulate is an easier place to look for quantile percentage. You drag the analysis variable(s) to the drop space for columns, drag the quantile statistic on top of Sum (the default) and type in the specific quantile). You can drag grouping variables to the left of the quantiles.