# Data Management-non-linear regression

• Apr 29th 2008, 05:02 AM
ives
Data Management-non-linear regression
To get a best fine line (non-linear ones) for a scatter plot,
what is the difference between a exponential curve and power regression curve?
These 2 curves are tooooo similar, i just can't choose one among them to fit the scatter plot```
can somebody give me an example or shapes of the scatter plots to distinguish them?
• Apr 29th 2008, 05:23 AM
TKHunny
For a given regression on specific data:
1) The Exponential will climb faster than the power in one direction.
2) The Exponential levels off in the other direction.

Over a very narrow Domain, they can be nearly indistinguishable. If you promise to stay in the very narrow Domain, it may not matter which you pick. If you attempt to extrapolate (outside the narrow Domain), you must be very, very careful.
• Apr 29th 2008, 05:29 AM
ives
Quote:

Originally Posted by TKHunny
For a given regression on specific data:
1) The Exponential will climb faster than the power in one direction.
2) The Exponential levels off in the other direction.

Over a very narrow Domain, they can be nearly indistinguishable. If you promise to stay in the very narrow Domain, it may not matter which you pick. If you attempt to extrapolate (outside the narrow Domain), you must be very, very careful.

-yea .. i forget.. the power only for one side of the y axis
usually i'll choose the exponential , cuz the curve looks better (Itwasntme)
than you
• Apr 29th 2008, 06:48 AM
TKHunny
There is much to be said for "looks better", but it is even better to have some sort of idea before you start. If you have a theoretical basis for a power, rather than an exponential, please do not use the exponential, even if they appear to be indentical.