Does anyone know how to fix the problem?
PS.Data
Moderators: phlip, Moderators General, Prelates
Code: Select all
f(x) = a + b*x + c*x**2 + d*x**3 + e*x**4 + f*x**5 + g*x**6
fit f(x) 'data' via a,b,c,d,e,f,g
plot f(x), 'data'
I needed to extrapolate intermediate values for some collected data, so I graphed it in Excel (scatter graph) and fitted a 6th order Polynomial to the data (damn good fit) I planned to nick the equation for the polynomial to predict values between my data points... Problem is that it doesn't work, the stated equation for the polynomial doesn't fit by several orders of magnitude!!
Does anyone know how to fix the problem?
PS.Data
Spoiler:Spoiler:
CrazyIvan wrote:A good rule of thumb is you should only at max one explanatory variable for every 10 observations you have. Which means really, your data can manage a linear fit, and a second order fit is stretching.
CrazyIvan wrote: I've also never, ever seen someone give a plausible reason for a 6-order term besides "well, it fit well". Which is *never* a reason to include it in the model.
mouseposture wrote:"It fit well" is a fine reason if, like the OP, all you want to do is extrapolate* the data, not understand it.
(But only ceteris paribus.)
*Wrote extrapolate, but apparently means interpolate.
CrazyIvan wrote:
Parsimony and model building is admittedly less of a big deal with a purely predictive model, but it still leaves a bad taste in my mouth. Doing anything with your data without a reason why gives me an uncomfortableness.
Users browsing this forum: No registered users and 23 guests