### Monday, March 06, 2006

## Re: st: Total Least Squares Regression, anyone

Thank you for the comment,

I already had, naturally, googled total least squares regression and had go= t a fair understanding what it is about, but no clue how to commence with State. Unfortunately I do not find singular value decomposition very useful= =2E

The statistical, rather than software, part of the problem is that the ANSI= , American National Standard, is vague in what exactly should be done.

The data are measurements with two instruments, one returning the reference value and the other, the test value. Ideally, both should be the same. The standard specifies that subtracting the latter from the former, the difference should be within defined tolerance limits 95% of the time. That part is easy to do. Then the standard states another requirement:

"The data, taken in the course of design compliance testing, should be analysed using the method of Total Least Squares Regression, and the slope of the regression, the offset of the regression and the standard deviation of the regression should be made available."

As I understand it, because both the reference and the test instrument are biased, total least squares regression should be used to take into account the uncertainty in both the dependent and the independent variable, and the relationship should be linear.

So my question remains, can Stata produce this output from the data I have. There is not reference in Stata to this method that I can find.

Best regards, Tero

-----Alkuper=E4inen viesti----- L=E4hett=E4j=E4: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Puolesta Marcello Pagano L=E4hetetty: 4. maaliskuuta 2006 18:55 Vastaanottaja: statalist@hsphsun2.harvard.edu Aihe: Re: st: Total Least Squares Regression, anyone

I don't know what the ANSI standard is referring to, but yes Stata should be helpful. Search for singular value decomposition within Stata. If you don't know the link with total least squares regression do a search on that phrase in Google.

m.p.

Tero Kivel=E4 wrote:

>Dear Statlisters, > >To fulfil the requirements of standard ANSI Z80.10-2003 I need to perform >"Total Least Squares Regression", not referenced or specified. > >Is this something that Stata can perform? I found nothing with the search >untility. If so, how? > >Thanks is advance > >Tero > > * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

* * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

Tag: statalist