regression - Microsoft Excel. Piece-wise Least-Squares Fit with Solver. My excel sheet produces right answers sometimes, wrong answers other times -

regression - Microsoft Excel. Piece-wise Least-Squares Fit with Solver. My excel sheet produces right answers sometimes, wrong answers other times -

i trying non-linear regression info have research. since nonlinear, can't utilize simplex lp. instead doing grg nonlinear upper , lower bounds on parameters.

it weird because excel produces answers right sometimes, , other times wrong. have manually alter parameters arbitrary numbers, run solver again, , hope right. allow me show excel sheet.

https://drive.google.com/a/case.edu/file/d/0bw0ajv0lw2etahfrufhobvz4nws/edit?usp=sharing

basically, looking 2 linear lines fit info have. raw info can divided 2 portions, both linear. point 2 lines cross critical value.

the right output raw info critical value = 0.006707. lastly time ran it, on excel sheet, can see critical value = 3.36e-06.

if help understand improve in context, measuring surface tension of various systems. critical value called critical micelle concentration in field.

thanks guys.

if using excel solver, think falling in local optima. restarting solver after setting different variables random values may help getting out of local optima values. before restarting solver, record values of objective function , compare new value found solver. alternatively, utilize improve solver outside excel. there plenty of open source solvers available on internet.

excel regression