- #1
PhysJs
- 6
- 1
Homework Statement
What is the error of the y-intercept and gradient of a fitted line? The line is fitted to some number of data points each with a known error in the x and y values.
More specifically, I have fitted a linear equation to some data points using Microsoft Excel LINEST, and the error given does not seem satisfactory.
Homework Equations
As far as I know Excel uses the method of least squares.
The specific procedure used in Excel is:
1) Select 4 cells
2) Use =LINEST([],[],1,1)
3)Press shift+ctrl+enter, 4 values pop out, 1 for y-intercept, 1 for x -intercept, and a cell for an error of each the y-intercept and x-intercept
The Attempt at a Solution
I have fitted a linear equation to some data points using Microsoft Excel LINEST function according to specified procedure. Only the values of x,y have to be given, Exccel doesn't seem to need to know the error in the data points.
Excel gives an 'error', but it seems to me that the given error does not depend on the errors of the data points, which seems counter intuitive. Surely if the errors on my data points were greater, the error in my gradient could be more. Thinking in pictures, if my data points had wider error bars in the y-direction, the linear line would be able to angle up and down with freedom, showing potential value the gradient can take.
This is confusing, not sure if I can trust Excel entirely. Is it somekind of mathematical reasoning that eliminates the dependence of the calculated gradient and y-intercept from the errors in the data points?