# Trend line equation

Posted by Denise G on January 03, 2002 1:10 PM

I've added a 6th order polynomial trend line to a line graph, and the fit is exactly what I'm looking for. Now, I'd like to use the polynomial equation to perform data analysis, but for some reason when I apply the equation the "y" values aren't anywhere close to where they should be. Here is the equation:

y = -8E-11x6 + 1E-08x5 - 1E-06x4 + 4E-05x3 - 0.0007x2 + 0.0104x + 0.0344

What could I be doing wrong???

Thanks.

Denise

Posted by Joe Was on January 03, 2002 9:19 PM

Try playing with the operation order to force Excel to perform the calculations as you intend, not at it wants to. Excel uses Addition then multiplication then powers, so group your operations with ( ). JSW

Posted by Bariloche on January 03, 2002 11:27 PM

Denise,

If your dataset is small enough could you post it so that we can have a go at it? Maybe post it even if it isn't small (just as long as its not like 3,000 points or so. LOL)

thanks

Posted by Denise G on January 04, 2002 7:08 AM

I would really appreciate it if you gave it a try -- Should I just copy the two columns of data (approx. 100 points) into this comment window?

Denise ,

Posted by Denise G on January 04, 2002 7:11 AM

Joe,

Tried it. Didn't work. Any other suggestions?

Denise

Posted by Denise G on January 04, 2002 10:21 AM

I think I've figured out the problem -- the \$ values may be interpreted as labels instead of values in the graph. I did not use a uniform scale in generating the sample data, and the scale does not appear to have been adjusted in the excel graph.

I don't know how to fix this, but progress has been made.

Denise

,

Posted by Bariloche on January 04, 2002 4:48 PM

post the data anyway

Denise,

Unless you've completely figured out the problem, you can post the data anyway. Ya never know what someone here might be able to figure out. And, yes, just pasting it in to the comment box should be ok.

Not sure exactly about your "scale" comment. To my way of thinking, anything you could do regrding the scale should not affect the equation of the fit curve.

enjoy

I think I've figured out the problem -- the \$ values may be interpreted as labels instead of values in the graph. I did not use a uniform scale in generating the sample data, and the scale does not appear to have been adjusted in the excel graph. I don't know how to fix this, but progress has been made.