Comments on: Trendline Fitting Errors

By: Jon Peltier

Jon Peltier — Fri, 03 Feb 2012 18:29:59 +0000

Tim -

Does deleting then redisplaying the label with the formula show an updated formula? If not, you probably need to delete and recreate the trendline itself with formula.

You can calculate the t-statistics for LINEST coefficients from the LINEST output table, and you can get p-values from these.

By: tim mayer

tim mayer — Fri, 03 Feb 2012 18:01:05 +0000

Hi Jon,

I have 2 questions: I have a series of graphs of climate data and stats for a given site. Each time I import data from a new climate station, I run a macro to calculate some seasonal climate statistics and then update the underlying data/titles for the graphs. The graphs are updated automatically (with one exception). I notice that when I change the underlying data, the trendline on the graph updates but the trendline formula on the graph does not. Is there a way to get the formula on the graph to update as well (I am using Excel 2010). Second, is there a way to quickly produce a p-value related to the slope of a linear trend (something alluded to by the first post here). I would like this to be autocomputed and shown on the updated graph as well. I know I can get a t-stat from the TINV fxn – maybe I just say something about the significance based on this?

By: ike9898

ike9898 — Tue, 10 Jan 2012 23:26:55 +0000

This article really helped me, specifically the part about using too few significant digits when copying down the coefficients for a fitted curve. Before I found this page, I was really stumped.

By: Jess Hughes

Jess Hughes — Wed, 26 Oct 2011 12:37:33 +0000

Hey great website, I’m having trouble with excel and getting mixed messages from the different people I ask or talk to. See the following forum where I have made a post about the problem I’m having and please let me know whether I have a good reason to be going crazy or whether I’m expecting too much of excel or whether I’m expecting too much of myself because I don’t know enough!

http://www.mrexcel.com/forum/showthread.php?t=587954

Cheers

Jesse

By: uday kumar

uday kumar — Tue, 03 May 2011 07:05:15 +0000

Many many thanks to you jon for answering expeditiously.
Actually, as you have written , the area can be calculated by trapeziod rule also; i knew this. But i got interest in trendline for two reasons- first, i saw that it fit the data very well (R^2=0.9996), so instead of calculating area using trapeziod rule, which is relatively more cumbersome than definite integral of a polynomial, I was tempted to use trendline equation; second- when trendline equation did not give the same y value or even near to actual y value then i was intrigued that why did it happen if it fitting so well.

As you have suggested in this post, I increased the precision of co-coefficients but even then i did not get the value near to actual value. Then i thought there is some problem with equation itself and i raised question on the equation.
Now when you plotted the fitted value using the trendline equation it is very much matching. at least it is not as far as i was getting.
THE REASON FOR THIS DISCREPANCY WAS THAT I DID NOT PROPERLY DRAG THE TRENDLINE EQUATION OVER ALL DATA POINTS. so the fitted y was wrong. -:)

sometimes people are just habitual to handle big problems but can not silly problems!

Regards
uday

–

By: Jon Peltier

Jon Peltier — Mon, 02 May 2011 05:02:48 +0000

Uday - I've worked through the exercise in Area Under a Fitted Curve.