# What I want is the uncertainty in the coefficients

1. Jul 23, 2009

### Winzer

1. The problem statement, all variables and given/known data
I was given a set of data from a Ion engine experiment. I fitted the data with an 8th order polynomial that seems pretty reasonable. What I want is the uncertainty in the coefficients.
One way I read was through the design matrix.

2. Relevant equations
$$C_{ij}=(A^T A)^{-1}$$

3. The attempt at a solution
The square root of the diagonals give the uncertainties. But my matrix in not invertible since its not square.
What am I missing?

2. Jul 23, 2009

### The Electrician

The matrix A may not be square, but

$$(A^T A)$$

is square, so it's invertible.

3. Jul 23, 2009

### Winzer

Thank you. Yes it turns out square but the inverse gives me problems.
Maybe I am using the command wrong.
Here are the results of a test.

I threw in .1 just for error sake. My model is a+bx+cx^2. The uncertainties are suppose to be given by the square root of the diagonals of
$$(A^T A)^{-1}$$ A being the design matrix.

#### Attached Files:

• ###### Screenshot.png
File size:
47.8 KB
Views:
63
4. Jul 24, 2009

### Winzer

Never mind I got it.
I was able to find the covariance matrix and the associated uncertainties. I feel stupid.