# Curve fitting when typical regression models fail

• Feb 3rd 2011, 12:20 PM
seligman28
Curve fitting when typical regression models fail
I tried using a regression model to fit a curve, but none of the typical models work. I'm not smart enough to use other regression methods, but I'm thinking an equation can be derived without it (hopefully).

The scatter plot shows Average Weight on the y-axis, and a ratio of the quantities of two objects, A and B, on the x-axis.

I know from work experience the average weight of object A is 33-34 tons, while object B is about 140-143 tons. The purpose of this is to estimate average tonnage by using the ratios of quantity A to quantity B.

Y-axis = Total Weight (various qtys of A & B combined) divided by the (quantity of A + quantity of B)

X-axis = quantity of A divided by quantity of B

Can someone help me with an equation to fit a curve?

Attachment 20679
• Feb 3rd 2011, 05:45 PM
dwsmith
Quote:

Originally Posted by seligman28
I tried using a regression model to fit a curve, but none of the typical models work. I'm not smart enough to use other regression methods, but I'm thinking an equation can be derived without it (hopefully).

The scatter plot shows Average Weight on the y-axis, and a ratio of the quantities of two objects, A and B, on the x-axis.

I know from work experience the average weight of object A is 33-34 tons, while object B is about 140-143 tons. The purpose of this is to estimate average tonnage by using the ratios of quantity A to quantity B.

Y-axis = Total Weight (various qtys of A & B combined) divided by the (quantity of A + quantity of B)

X-axis = quantity of A divided by quantity of B

Can someone help me with an equation to fit a curve?

Attachment 20679

You could try

$\displaystyle f(x)=\beta e^{-x+\alpha} \ \ \ \alpha,\beta\in\mathbb{R}$
• Feb 3rd 2011, 05:49 PM
Prove It
I don't see why a Least Squares Linear Regression would fail here...
• Feb 3rd 2011, 07:38 PM
CaptainBlack
Quote:

Originally Posted by seligman28
I tried using a regression model to fit a curve, but none of the typical models work. I'm not smart enough to use other regression methods, but I'm thinking an equation can be derived without it (hopefully).

The scatter plot shows Average Weight on the y-axis, and a ratio of the quantities of two objects, A and B, on the x-axis.

I know from work experience the average weight of object A is 33-34 tons, while object B is about 140-143 tons. The purpose of this is to estimate average tonnage by using the ratios of quantity A to quantity B.

Y-axis = Total Weight (various qtys of A & B combined) divided by the (quantity of A + quantity of B)

X-axis = quantity of A divided by quantity of B

Can someone help me with an equation to fit a curve?

Attachment 20679

Is this not the same question (but in disguise) as in your other thread?

CB