comparison of 2 values – significant
Six Sigma – iSixSigma › Forums › Old Forums › General › comparison of 2 values – significant
 This topic has 8 replies, 6 voices, and was last updated 14 years, 9 months ago by Rajib Chatterjee.

AuthorPosts

April 6, 2007 at 6:51 pm #46656
Hello,
Do you know of a test one would use to make a statement about how close a modeled value is to a known value?
I have a known mean value of 14.54 for soil carbon. I then used a model to estimate the soil carbon value and the result was 14.14
I would like to make a statement about how close these values are.
Thank you for your time and any information,
darcy0April 6, 2007 at 7:31 pm #154500do you want to compare two samples?
0April 6, 2007 at 7:34 pm #154501sorry i missinterprept your question
u might want tu use 1 sample t test.
compare one sample to a single value0April 6, 2007 at 7:41 pm #154502Not really 2 different samples, but a comparison between a measured value and a modeled value. They’re quite close so I’m happy about that and want to show that a model worked well with some kind of statement of significance… But perhaps there isn’t a statistical method for this. I could just make a statement like “results show good agreement between measured and modeled values”.
Thank you,
darcy0April 7, 2007 at 7:49 am #154513
Adrian P. SmithParticipant@AdrianP.Smith Include @AdrianP.Smith in your post and this person will
be notified via email.Hi Darcy
I think what you would have to do is to collect mulitple soil carbon values and then model each one.
You could then do a paired ttest to demonstrate that the difference between the measured and modeled values is not significant.
Rgds,Adrian
http://adrian.smith.name
0April 9, 2007 at 3:59 pm #154557
Robert ButlerParticipant@rbutler Include @rbutler in your post and this person will
be notified via email.Your use of the term “modeled value” suggest you have some kind of output (prediction) from a regression equation. If this is the case then that equation will have an associated root mean square error (RMSE). A very quick and dirty common practice is to take your predicted value (the modeled value) and a range of + – 2 times the RMSE around that predicted value (roughly the 95% CI) and see if your measured response falls inside those limits. If it does then you can say that your measured and your modeled values are in agreement.
0April 9, 2007 at 4:30 pm #154560Thank you, I can try this.
0April 9, 2007 at 4:49 pm #154562
Jim ShelorParticipant@JimShelor Include @JimShelor in your post and this person will
be notified via email.I think Adrain’s suggestion is a very good one.
I would like to add one comment. Make sure your range of samples covers you range of interest. That way you make sure your model works across the full range with no linearity or discontinuity issues.0April 12, 2007 at 6:38 am #154691
Rajib ChatterjeeParticipant@RajibChatterjee Include @RajibChatterjee in your post and this person will
be notified via email.I have faced the same kind of problem. I got a value from the Testing machine & one from prediction model. Both are having some difference.
Prediction Data & the Data which we are getting from any testing machine could not be same, due to the tolerance level of each procedure. Only thing which we can do is to minimize the difference between the two data.
Solution : In my model, I have put a constant (k) which is nothing but a multiplication factor. Now while comparing the data which are from different source I use “Excel Solver” to minimize the difference between the two data.
This gives me a very good result.
Previously the difference was 0.159 or more, now it is come down to 0.0032.
Regards
Rajib Chatterjee
Jamshedpur India
[email protected]
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.