|
Data Gallery: Guenther/ISO NOx Calibration (Errors-in-Variables Regression) |
|
| Introduction |
Frank Guenther is head of CSTL's Gas Metrology program in the
Analytical Chemistry Division. The NIST Gas Metrology calibration
service plays a critical role in the measurement of a wide
spectrum of gases worldwide, and is a major player in ISO
metrology.
The data analyzed below comes from a Guenther run involving the measurment of nitrous oxides. The calibration problem is complicated by the fact that the horizontal axis also has (non-negligible) noise, and so usual regression techniques do not apply. The statistical classification of this problem is "Errors-in-Variables". To show the flexibility of the Bayesian approach, we focus on the aspect of estimating the slope from the errors-in-variables linear fit. Such slope calculation (along with the intercept calculation) is a necessary intermediate step to the computation of the final calibration values. You may also download the Guenther data. |
Data
|
||
Prior Distribution 1:
|
Prior Distribution 2:
|
Prior Distribution 3:
|
Posterior Distribution 1
|
Posterior Distribution 2
|
Posterior Distribution 3
|
Posterior Statistics 1
|
Posterior Statistics 2
|
Posterior Statistics 3
|
|
Date created: 9/21/2001 |
||
| [ SED Home | Bayesian Home | Data Gallery | Previous | Next ] |