Minos() ignores 'errors'

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Run Migrad()
2. Change a value in the 'errors' dictionary that migrad() spits out
3. Run Minos() with the altered 'errors'
4. Minos() uses the original error as an initial guess instead of the changed 
value

What is the expected output? What do you see instead?

After running migrad() I set a new value for one of the parameter's errors like 
so,   
m.errors['b'] = (.03)

When I run m.minos('b',1) I expect to see the first iteration on parameter 'b' 
increase by .03 but instead it increases by the original error value.

I feel that minos() grabs the error straight from the covariance matrix which I 
am not able to alter.  Is there a way of changing the initial guess for minos() 
because the pyminuit reference says minos() reads the initial guess from the 
'errors' dictionary.

Original issue reported on code.google.com by kcremin%[email protected] on 24 Aug 2011 at 6:41