Calibration, uncertainty, smt else (Paul) - Derived uncertainty

10 important questions on Calibration, uncertainty, smt else (Paul) - Derived uncertainty

THere's a relation between T and Qdrain, but we do not know its exact form

Finding probability might be too expensive/difficult

When is the (un)certainty covariance plot an elipse?

When there is a strong (positive or negative) correlation between the derived quantities or parameters.
When the covplot is nearly a circle, there's hardly any correlation

What is the order of the fitting parameter, deriving quantity and getting their uncertainties procedure?

1. Parameterise (give first values of all parameters)
2. Plot the model
3. Add observations
4. Plot model + observations
5. Fit parameters, eg 2
6. Plot model + fitted parameters > gives better representation based on observations
7. Get stats on fitted parameters (available in R), which gives uncertainty measurse like sd/var/mean which you can plot into a boxplot
8. Get/choose derived quantities, using the fitted parameters
9. Calculate slope of quantity wrt parameter
10. Get general stats and plot them into boxplot and covplot
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

Why is slope calculation necessary within this process?

?? Something about the approximately true and the approximation of the derived uncertainty of fitted T. More specifically: to calculate the variances or the covar matrix

What happens to uncertainty of derived quantity when plotting with more fitted parameters?

Fitting more parameters: uncertainty decreases (a little bit).

What is the Gaussian escape?

1. As probability calculation is very difficult (and expensive)
2. We assume f is linear, thus f(x) = ax+b
3. Thus: if E[x] and VAR[X] known, if f is linear, if Y is gaussian, then our problem is solved

Explain the Gaussian escape for derived uncertainty example

X = T, Y = Qdrain

  1. mean(T) and var(T) are known from fitting procedure
  2. F is linear - assume approximately true
  3. Y is Gaussian - assume approx true


Problem is approximately solved

Why not use a Latin Hypercube Sampling?

  • Works only for independent variables
  • Parameters are almost never independent after fitting

What is the covar of a linear function?

COV(y) = A*COV(x)*Atransposed

How is fitting done? Write down the formula

Finding min values of SS:

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo