CODEWITHSUNDEEP
r
user1471980
user1471980

Reputation: 10626

Is it possible to print out the model formual in R

Is it possible to print out the mathematical formula of the model in R. For example,

library(usl)

data(raytracer)
usl.model <- usl(throughput ~ processors, data = raytracer)

I would like to be able to look at this model's (usl.model) formula. Any ideas wheter is this doable or not?

Upvotes: 0

Views: 582

Answers (2)

Metamorphic
Metamorphic

Reputation: 791

The 'usl' documentation says:

The Universal Scalability Law can be expressed with following
formula. ‘C(N)’ predicts the relative capacity of the system for a
given load ‘N’:

   C(N) = N / (1 + sigma * (N - 1) + kappa * N * (N - 1))

I'm not sure how to interpret your question, but would it work to take the sigma and kappa coefficients from summary(usl.model) and plug them into this equation?

Update: For the raytracer example, it looks like N represents the number of processors and C(N) the throughput:

N=raytracer$processors;
attach(as.list(usl.model@coefficients))
[email protected]
data.frame(raytracer, pred=F* N / (1 + sigma * (N - 1) + kappa * N * (N - 1)))

This produces:

   processors throughput       pred
1           1         20  20.00000
2           4         78  69.55746
3           8        130 118.48067
4          12        170 154.75037
5          16        190 182.70168
6          20        200 204.89242
7          24        210 222.92923
8          28        230 237.87214
9          32        260 250.44877
10         48        280 285.56580
11         64        310 306.91823

I'm not sure how sigma and kappa are calculated. The documentation says that the default method uses "a transformation into a 2nd degree polynom" (citing Gunther) but if the model frame lacks a predictor with value 1 then it seems to fall back to optimizing the coefficients (scale factor, sigma and kappa) to minimize the sum of squared differences between the actual and predicted response variable (the last two columns above).

Upvotes: 1

Vincent Bonhomme
Vincent Bonhomme

Reputation: 7443

It is in the call slot of the model. You can access it with:

> usl.model@call
usl(formula = throughput ~ processors, data = raytracer)

A more compact option, eg for display purpose, is:

> print(as.formula(usl.model@call), showEnv=FALSE)
throughput ~ processors

More generally, on complex objects/lists/etc., str() is a nice entry point on the structure.

> str(usl.model)
Formal class 'USL' [package "usl"] with 15 slots
..@ frame        :'data.frame': 11 obs. of  2 variables:
  .. ..$ throughput: num [1:11] 20 78 130 170 190 200 210 230 260 280 ...
.. ..$ processors: num [1:11] 1 4 8 12 16 20 24 28 32 48 ...
.. ..- attr(*, "terms")=Classes 'terms', 'formula' length 3 throughput ~ processors
[.../...]
..@ call         : language usl(formula = throughput ~ processors, data = raytracer)
..@ regr         : chr "processors"
..@ resp         : chr "throughput"

You can also add @ (or $) to the object name (eg usl.model@) and then press <tab>.

Upvotes: 1

Related Questions