The [MINos]MINOS algorithm is described in some detail in part 1 of this manual. Here we add some supplementary ``geometrical interpretation'' for the multidimensional case.
Let us consider that there are just two free parameters, and draw
the contour line connecting all points where the function takes on
the value
.
(The [CONtour]CONTOUR command will do this for you from Minuit).
For a linear problem, this contour line would be an
exact ellipse, the shape and orientation of which are described in
[5], p.196 (fig. 9.4).
For our problem let the contour be as in figure 
.
If [MINos]MINOS is requested to find the errors in parameter
one (the x-axis), it will find the extreme contour points A and B,
whose x-coordinates, relative to the x-coordinate at the minimum
(X), will be respectively the negative and positive [MINos]MINOS errors of
parameter one.
[MINos]MINOS errors for parameter 1