I've just been through my first lab in my college physics studies. The exact nature of the lab itself isn't relevant - however, the data analysis is.
In our case, we presumed a relationship like t = lαθβrγ for the behaviour of our system, with exponents constant. We varied each of these three parameters and put them into tables. So when plotting, we treated the other parameters as a constant thus t = vαC. Next, in order to find the exponent α, we plotted log(t) = log(vαC) = α * log(v) + log(C). Linear regression is performed, and α is read as the coefficient of the resulting linear equation. Rinse and repeat for every parameter, and, given well-constructed parameters, there's your equation.
Pretty straightforward.
But WHAT is this called? Our teacher refers to it consistently as "linearization" and as distinct from "regression". The thing is, when I look up this online, both in Swedish and English, "linearization" is a general method for estimating the values of non-linear functions using tangent equations. Regression is certainly the final step of the method, but I haven't found anything relevant on logarithmizing in articles on regression. I have found nothing about the method I describe above in my searches, and I'd like to read some discussion on it, limitations and so on.
Thanks.
In our case, we presumed a relationship like t = lαθβrγ for the behaviour of our system, with exponents constant. We varied each of these three parameters and put them into tables. So when plotting, we treated the other parameters as a constant thus t = vαC. Next, in order to find the exponent α, we plotted log(t) = log(vαC) = α * log(v) + log(C). Linear regression is performed, and α is read as the coefficient of the resulting linear equation. Rinse and repeat for every parameter, and, given well-constructed parameters, there's your equation.
Pretty straightforward.
But WHAT is this called? Our teacher refers to it consistently as "linearization" and as distinct from "regression". The thing is, when I look up this online, both in Swedish and English, "linearization" is a general method for estimating the values of non-linear functions using tangent equations. Regression is certainly the final step of the method, but I haven't found anything relevant on logarithmizing in articles on regression. I have found nothing about the method I describe above in my searches, and I'd like to read some discussion on it, limitations and so on.
Thanks.
via JREF Forum http://forums.randi.org/showthread.php?t=267846&goto=newpost
Aucun commentaire:
Enregistrer un commentaire