Trendline Types

[Dave]

I am new to trendlines.

I am using an xy scatter graph. Is there rule for when to use the different
types of trendlines?

I presently use a Power Type, but I am not sure if this is giving me a true
tend.

Thanks

 

[Jon Peltier]

Dave -

Do you have any physical basis to use one over another? In other words, if
it's a physical phenomenon you're plotting, is there some theoretical
relationship that describes the phenomenon? If so, use the trendline of that
form, or transform the data until the model matches one of the available
types.

If you want a nice-looking, smooth fit, you could use a polynomial, but
people usually take this to extremes. There is no physical basis for a sixth
order fit. You should use the simplest equation that shows a reasonable
agreement with your data.

If you are using the fit for interpolation, use your judgment. Try to avoid
using a fit for extrapolation, especially a poly fit.

- Jon

 

[James Silverton]

Jon wrote on Wed, 3 Oct 2007 09:16:59 -0400:

JP> Do you have any physical basis to use one over another? In
JP> other words, if it's a physical phenomenon you're plotting,
JP> is there some theoretical relationship that describes the
JP> phenomenon? If so, use the trendline of that form, or
JP> transform the data until the model matches one of the
JP> available types.

JP> If you want a nice-looking, smooth fit, you could use a
JP> polynomial, but people usually take this to extremes. There
JP> is no physical basis for a sixth order fit. You should use
JP> the simplest equation that shows a reasonable agreement
JP> with your data.

JP> If you are using the fit for interpolation, use your
JP> judgment. Try to avoid using a fit for extrapolation,
JP> especially a poly fit.

JP> - Jon
JP> -------
JP> Jon Peltier, Microsoft Excel MVP

JP> ??>> I am new to trendlines.
??>>
??>> I am using an xy scatter graph. Is there rule for when to
??>> use the different types of trendlines?
??>>
??>> I presently use a Power Type, but I am not sure if this is
??>> giving me a true tend.
??>>

Jon's post looks like excellent advice. I sometimes find a
moving average to be enlightening but I don't try to make future
predictions from it.

James Silverton
Potomac, Maryland

E-mail, with obvious alterations:
not.jim.silverton.at.verizon.not