Quadratic Regression - Help

**kmr159** · 04-07-2013, 10:09 PM

Background
I am trying to determine the value of g based on a simple pendulum. The equation, for a small angle approximation works out to L=[G/((2pi)^2)] P^2. Where L is the length of the Pendulum in meters, G is Gravity and P is the period of the pendulum. I have graphed a scatter plot with the x axis as L and the y axis as P.

I have 3 different series for different lengths of pendulum.
When I try to find a line of best fit I get something like ax^2 + bx + c and this trendline is only for one series.

Question
2 Parts - Is there any way to get the trendline to be the line of best fit for all my different series ( different lengths)
- Is there any way to make my quadratic regression approximate only in terms of ax^2.

Thanks
Physics Student

**zbor** · 04-08-2013, 02:13 AM

Can you upload example workbook with your inputs and desired output?
You can hide any private and professional information but keep what's important to understanding (excel and math) problem.

ax^2+bx+c can be replaced with ax^2 for big x.

**MrShorty** · 04-08-2013, 08:26 AM

- Is there any way to make my quadratic regression approximate only in terms of ax^2.

Consider this substitution:

Let u=x^2.
Substitute into y=ax^2 --> y=au. It should (hopefully) be obvious that y as a function of u is a straight line through the origin (y=au+0).
When I do this, I add a column that calculates u, then regress against that column.
Setting the third argument in the LINEST() function (http://office.microsoft.com/en-us/ex...005209155.aspx) to false should cause Excel to set b=0.