Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let's suppose we have a function $Y=A\cdot t^B$ and the values for $Y$ are $30,60,90,120,150$ and the values for $t$ are respectively $0.974, 1.331, 1.718, 1.971, 2.356$. Can you find $A$ and $B$ with the method of linear regression? I have to do a lab work and this is a very small part of it, which does not count but I still have to do it and I have never done linear regression, I need this now? please?

share|improve this question
    
can't you find A and B without regression? –  dato datuashvili Jan 4 '13 at 10:17
    
I need that with the regression method..that's what it says in my paper –  ddsd Jan 4 '13 at 10:18
    
then i think it is not linear regression model,it is called exponential regression –  dato datuashvili Jan 4 '13 at 10:19
    
in case of any question,please ask –  dato datuashvili Jan 4 '13 at 10:30
1  
I highly doubt that the pedagogical purpose of this question is to do power/exponential regression. A straight line can be fitted using linear least squares fit. For power/exponential regression, one has to use non-linear least squares fitting...which is much more non-trivial than LLSF. –  Fixed Point Jan 4 '13 at 10:55
show 3 more comments

2 Answers

Are you allowed to use a calculator or software or do you have to do this by hand? In either case, if you have $y=at^b$ then just take the logarithm of both sides and get $$\log(y)=\log(a)+b\log(t)$$ and then use linear regression using this form of the fitting function. So remember you won't put in your data points $y$ and $t$ as you have them above. You have to take their log and use $\log(y)$ and $\log(t)$. Then the slope of the fitted line will be your $b$ because that is the coefficient in front of $\log(t)$ and the y-intercept will be $\log(a)$ so exponentiate it to get $a$ and VOILA!

share|improve this answer
    
exactly,in fact there is no big difference if it is power or exponential,they have similar solution –  dato datuashvili Jan 4 '13 at 10:53
1  
Yep, they wouldn't be exactly the same but they are close and in this case both answers are within 95% confidence interval of each other. The power regression gives me $a=36.45$ and $b=1.679$. –  Fixed Point Jan 4 '13 at 11:18
    
In my results I get a=30 and b=2 ..the values of t and y are not very exact because of lab work ,and in my results i get different b's and a's ( sometimes b=1.7 ,b=2.1,b=2.2 )for different systems (of y and t) so i will take a=30 and b=2, i have to take exact values because i have to build a graph ..in this case the graph will look smth like Y=30*b^2 –  ddsd Jan 5 '13 at 10:56
add comment

http://www.zweigmedia.com/RealWorld/calctopic1/regression.html look here it contains both linear regression and exponential regression

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.