S.O.S. Mathematics CyberBoard

Hi,
I'm having some problems getting a model for some experimental data. I need a function of Air Flow with dependent variables Boost and RPM. Here's the data:

RPM Boost Air Flow
0 0 0
600 0 4
1200 0 7.8
1800 0 11.6
2400 0 14.4
3000 0 16.5
0 10 0
600 10 8.6
1200 10 16.75
1800 10 24.9
2400 10 27.4
3000 10 30
0 20 0
600 20 13.2
1200 20 25.7
1800 20 38.2
2400 20 40.5
3000 20 43.5
0 30 0
600 30 18.6
1200 30 35.2
1800 30 51.8
2400 30 55.05
3000 30 58.3

Can someone suggest a method and possibly where I could learn the method. Or, better yet, get me started on the method. I have a week to get a model for my boss with a 2%error max. I tried using the least square method with a linear function and got an average of 20% error.

My function for the LMS method was:
Air Flow = .0126*RPM+.9141*Boost-10.078

Any help would be much appreciated,
Tu Le

Have you tried simply plotting your data? The line reading

2400 20 10.5

looks fishy.

that's because its supposed to be 40.5 and not 10.5 ><
I did plot it and it appears linear until towards the end when it looks for like a y=sqrt(x) sort of curve.

My advice is to think about the physics of the experiment and think about what is a plausible form of the function. (I have no idea what you mean by Boost, for example.) You are likely to get better results that way than by treating the data as simply a pure set of numbers and looking for a "good" fit.