Instead of selecting Fit Model, one may also use Graph - Variability/Gage
Chart. Select your X and Y and then under Variability Gage, select Variance
Component.
The advantages are that you get the REML method and you get the Variability
Chart (which helps you visualize the within lot versus the between lot
variability). See the .jrn file.
Karl
_________________________________________________
-----Original Message-----
From: jmp-l-owner@lists.biostat.wustl.edu
[mailto:jmp-l-owner@lists.biostat.wustl.edu] On Behalf Of
Gunter.Hartel@csl.com.au
Sent: Tuesday, August 30, 2005 6:31 PM
To: jmp-l@lists.biostat.wustl.edu
Subject: Re: [jmp-l] Simple Variance Component Analysis
Dear Sean,
That can happen sometimes when the within lot variance is actually larger than
the between lot variance. For example in biometrics you can have a within plot
negative correlation arising from a competitive situation where one plant being
larger makes the other plant in that plot smaller.
Looking at your data you have a situation where in some lots the difference is
extremely small, eg sd<10, and in some lots it is quite large, eg sd>300. You
might see if there is an explanation for that.
Cheers
Gunter
-----Original Message-----
From: jmp-l-owner@lists.biostat.wustl.edu
[mailto:jmp-l-owner@lists.biostat.wustl.edu] On Behalf Of Davern, Sean
Sent: Wednesday, 31 August 2005 6:13 AM
To: 'jmp-l@lists.biostat.wustl.edu'
Subject: Re: [jmp-l] Simple Variance Component Analysis
So what does it mean when the Lot&Random component of variance
is -22% of
the total?
How is it possible for the Var Comp Est to be negative?
Sean
-----Original Message-----
From: jmp-l-owner@lists.biostat.wustl.edu
[mailto:jmp-l-owner@lists.biostat.wustl.edu]On Behalf Of
Martin_Kane@hgsi.com
Sent: Tuesday, August 30, 2005 12:51 PM
To: jmp-l@lists.biostat.wustl.edu
Subject: Re: [jmp-l] Simple Variance Component Analysis
The correct way to do this analysis is to use the "Fit Model"
analysis.
Put "Response" in the Y role and put "Lot" in the Effects role.
Now,
highlight "Lot" in the effects box and select the red
"Attributes" arrow.
With this "Attributes" menu, select "Random Effect". This will
change the
effect from "Lot" to "Lot& Random".
Doing this will allow you to get variance components out.
Unfortunately, your data set seems to have much more
variability within lot
than lot-to-lot.
Sincerely,
Martin Kane
Manager, Non-Clinical Statistics
Human Genome Sciences, Inc.
240-314-4400 x3822
"Davern, Sean"
<sdavern@amgen.com> To:
"'jmp-l@lists.biostat.wustl.edu'"
Sent by:
<jmp-l@lists.biostat.wustl.edu>
jmp-l-owner@lists.biosta cc:
t.wustl.edu Subject:
[jmp-l]
Simple Variance Component Analysis
08/30/2005 03:42 PM
Please respond to jmp-l
Can anyone tell me how to get JMP to give a very simply
variance component
analysis?
I'm trying to quantify lot-to-lot variability of a raw
material. I've made
two independent measurements of 9 lots. I'd like to segregate
the overall
variance into within lot variation (coming from run-to-run
variation) and
between lot (lot-to-lot) variation. I'd like the total
variance to be
reported as the two components.
Attached is the example data set.
<<Example Data.JMP>>
It seems like this should be very simple but I can't seem to
get the
answer without doing the calculations by hand.
Thanks,
Sean
Sean Davern
Engineer III
Cell Sciences Process Development
Mail Stop AW2/D2152
Ext. 57074
(See attached file: Example Data.JMP)
example data1.JRN
Description: Binary data
|