In doing basic linear regression on graphing calculators, (given a list of data points, find a best fit line) there are occasional differences between the equation given by a TI calculator and that given by a Casio calculator.
We know that TI’s use least squares to find the best fit line. Anyone know how Casios do? Our best bet is that the Casio somehow ignores outlying points.
I don’t know. But can I ask you something? Why are you using a calculator to perform linear regression? Is it part of a school assignment? Because given the choice, I would much rather use Excel…
Least squares is pretty standard. The other option that I can think of is to minimize the sum of the absolute values of the distances; that wasn’t widely used before computers because it’s a bitch to do except numerically.
The other possibility is that floating point error is causing trouble for one of them. How large is the difference relative to the size of the coefficients?
And SAS is by far the best tool for linear regression, IMO. Excel is good for really basic stuff, but has some severe limitations.