The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

Understanding polynomial evaluation of linearity

Vishesh Shrivastav

2019-02-21

lin.eval

R package for polynomial evaluation of linearity.

About

lin.eval is a R package for performing polynomial evaluation of linearity.

How it works

Polynomial evaluation of linearity is a technique of assessing if the best way to describe the relationship between two vectors.

  1. Fit three models - linear, second-order polynomial and third-order polynomial
  2. Find out best-fitting model among the three by comparing their p-values. Model with the lowest p-value out of the three is the best-fitting one.
  3. If the best-fitting model is linear, linearity is established and no further steps need to be carried out. This is called Linear 1 type.
  4. Else, best-fitting model is either second or third order polynomoal model. In this case, calculate average deviation from linearity (adl). This is given by:

\[adl\ =\ \frac{1}{n} * (\sum_{1}^{n}|\frac{l_i - p_i}{l_i}| * 100)\]

where, l is the vector of predictions from linear model and p is the vector of predictions from best-fitting polynomial model.

  1. If adl is greater than or equal to the threshold value for deviation from linearity, conclude that the relationship is non-linear.
  2. Else if adl is less than the threshold value for deviation from linearity, conclude that although the best-fitting model is not linear, deviation from linearity is not significant and hence, it is still a linear relationship. This is called a Linear 2 type.

Usage

Call the poly_eval() function with the following parameters:
y: vector of response values
x: vector of predictor values
threshold: threshold value for average deviation from linearity as percentage. Defaults to 5.

library(lin.eval)
foo <- c(165.3929, 165.3929, 1119.5714, 1119.5714, 2073.7500, 2073.7500, 3027.9286, 3027.9286, 3982.1071, 3982.1071, 4936.2857, 4936.2857, 5890.4643, 5890.4643)
bar <- c(386.2143,  386.2143, 840.6548, 840.6548, 1829.6905, 1829.6905, 3074.4048, 3074.4048, 4295.8810, 4295.8810, 5215.2024, 5215.2024, 5553.4524, 5553.4524)
derp <- poly_eval(bar, foo, 30)
#> Best fitting model is third-order polynomial.
#> Computing average deviation from linearity:
#> Average Deviation from Linearity: 27.28 %
#> Although the best fitting model is nonlinear, since average deviation from linearity is 27.28; which is less than or equal to 30; linearity is established. We call this linearity type as Linear 2

You can check the values stored in the result variable:

derp$p1
#> [1] 8.851095e-12
derp$p2
#> [1] 2.514044e-10
derp$p3
#> [1] 1.930392e-78

More examples

Usage without passing in optional argument for adl:

foo <- c(0, 1, 2, 4, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 19, 20, 21, 22, 24, 25, 26, 27, 28, 29, 30)
bar <- c(126.6, 101.8, 71.6, 101.6, 68.1, 62.9, 45.5, 41.9, 46.3, 34.1, 38.2, 41.7, 24.7, 41.5, 36.6, 19.6, 22.8, 29.6, 23.5, 15.3, 13.4, 26.8, 9.8, 18.8, 25.9, 19.3)
poly_eval(bar, foo)
#> Best fitting model is second-order polynomial.
#> Computing average deviation from linearity...
#> Average Deviation from Linearity: 70.42 %
#> Since, average deviation from linearity is greater than 5, nonlinearity is established.
#> The relationship between the two input vectors is best described by a second order polynomial

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.