Greetings.
I have two columns with data in LO Calc, one representing years and one representing population. How do I perform a quartic regression on this data? My model for the regression is ax⁴+bx³+cx²+dx+f.
I’m running the Flatpack distribution of LibreOffice on Debian GNU/Linux 11 (bullseye).
Thank you, and God bless.
Study and try the array function LINEST()
@Zizi64 is right concerning polynomial approximations. LINEST() can be used to get them.
Regard the slightly strange fact that the coefficients of the polynomial and the exponents to be used to create the columns with the powers of the original variable (mostly named “x”, in your case taking the year values) are ordered reversely…
Anyway I am interested in a realistic example of your data and in the reasons for the expectation to get a useful model based on polynomials. Polynomials approach infinity on both sides, after all.
Likely a (regional?) convention. The order (largest powers come first) looks perfectly familiar to me (who had been taught in a Russian school ;))
and to me (who was educated in Germany) too.
(Thanks for the opportunity to try to read a Russian text. Last time I did with partial success was in 1969. It was from a mathematical magazine and treated questions concerning topological projective planes.)
However, your remark may be based on a misunderstanding induced by my badly worded post. It referred to the example (polynomials) while it actually addressed multilinear approximations (what’s also behind the fitting polynomials) in general. The order of the returned coefficients is reverse as compared with the order of the related columns.
See
disaskl80726PolynomialApproxAndChart.ods (37.6 KB)
