Exponential Regression
- Thread starter shanepby
- Start date
- Joined
- Feb 4, 2004
- Messages
- 16,583
As near as I can tell (from Googling the topic), you use logs to convert the data to something linear, find the linear regression, and then work backwards to the exponential formula. But even the linear formulas are fairly complex....
Eliz.
Eliz.
John_Bilout
New member
- Joined
- Aug 19, 2005
- Messages
- 11
Determine how the constants a & b are determined in exponential regression, y=b*a^x and explain why the formulas work. Thanks in advance for any help.
Well, i think u should turn a^x into e^(lna *x) then the solution is there : the formulas work only if a >0 ( definition domain of ln ).
i hope it'll help u...
Well, i think u should turn a^x into e^(lna *x) then the solution is there : the formulas work only if a >0 ( definition domain of ln ).
i hope it'll help u...
- Joined
- Apr 12, 2005
- Messages
- 11,339
y = b*e<sup>a*x</sup>
Transform with Logartihms
log(y) = log(b) + x*log(a)
This is now linear in (x,log(y)) and Normal equations can be constructed without too much dificulty. You will need:
Count(data points)
Sum(x values)
Sum(x<sup>2</sup> values)
Sum(log(y) values)
Sum(x*log(y) values)
It should be obvious that y > 0. If this is not the case, some modification will be required.
Transform with Logartihms
log(y) = log(b) + x*log(a)
This is now linear in (x,log(y)) and Normal equations can be constructed without too much dificulty. You will need:
Count(data points)
Sum(x values)
Sum(x<sup>2</sup> values)
Sum(log(y) values)
Sum(x*log(y) values)
It should be obvious that y > 0. If this is not the case, some modification will be required.