sim.measure-class: Similarity measure

In AlmaasLab/micInt: Find microbial interactions

Similarity measure

Description

A similarity measure structure with the function ifself, and metadata fields

Usage

sim.measure(
  FUN,
  string,
  categorical = FALSE,
  mean_scaleable = FALSE,
  signed = FALSE,
  type = "parametric"
)

Arguments

FUN	The similarity measure function to use, of the form f(x,y). The function must satisfy the following: If x and y are both given, they are treated as vectors and the similarity score between them is returned If only x is given, it is treated as a matrix (or data frame) where the features are in columns and sample in the rows. The function must then return the matrix of pairwise similarities The function must return a similarity score in the interval [0,1], where 1 is perfect similarity and 0 is perfect dissimilarity. Alternativly, signed similarity scores in the interval [-1,1] can also be used.
string	The similarity score's human readable name.
categorical	Logical value indicating whether the function only considers presence-absence of an OTU. If TRUE, the function will not be noisified as there is no point in adding noise to the data.
mean_scaleable	Logical value telling whether it makes sense to scale the input by its mean before applying this function
signed	If TRUE, the similarity measures gives signed results in the interval [-1,1], else they are in the interval [0,1].
type	Character; the way the similarity measure processes the information. On of: "presence-absence" Measures the similarity of OTUs solely by their presence-absence pattern "non-parametric" Measures the similarity based on the rank pattern, but besides this, no other information on the abudandances are used "parameteric" The actual values of the abundances matter in the calulations

Slots

FUN

The similarity measure function to use, of the form f(x,y). The function must satisfy the following:

• If x and y are both given, they are treated as vectors and the similarity score between them is returned

• If only x is given, it is treated as a matrix (or data frame) where the features are in columns and sample in the rows. The function must then return the matrix of pairwise similarities

• The function must return a similarity score in the interval [0,1], where 1 is perfect similarity and 0 is perfect dissimilarity. Alternativly, signed similarity scores in the interval [-1,1] can also be used.

string

The similarity score's human readable name.

categorical

Logical value indicating whether the function only considers presence-absence of an OTU. If TRUE, the function will not be noisified as there is no point in adding noise to the data.

mean_scaleable

Logical value telling whether it makes sense to scale the input by its mean before applying this function

signed

If TRUE, the similarity measures gives signed results in the interval [-1,1], else they are in the interval [0,1].

type

Character; the way the similarity measure processes the information. On of:

• "presence-absence" Measures the similarity of OTUs solely by their presence-absence pattern #'

• "non-parametric" Measures the similarity based on the rank pattern, but besides this, no other information on the abudandances are used

• "parameteric" The actual values of the abundances matter in the calulations

See Also

noisify

Examples

# This measures how well the arguments agree with the signs
# This primary function accept two vectors only
my_measure_simple <- function(x,y){
 (sum((x > 0) & (y > 0)) + sum((x < 0) & (y < 0))) /
  (sum(x > 0) + sum(x < 0))
}
# We extend this function into acception a matrix argument when y is now supplied
my_measure <- function(x,y=NULL){
if(is.null(y)){
n_vectors <- ncol(x)
res <- matrix(NA_real_,n_vectors,n_vectors)
for (i in seq_len(n_vectors)){
    for(j in seq_len(n_vectors)){
    res[i,j] <- my_measure_simple(x[,i],x[,j])
    }
}
    res

}
else{
my_measure_simple(x,y)
}
}

my_sim_score <-sim.measure(my_measure,string= "My measure", categorical = FALSE,
 mean_scaleable = FALSE, signed = FALSE, type = "non-paramtric")
sim_fun <- sim_measure_function(my_sim_score)
sim_fun(c(1,0,-1),c(2,-1,-1))
sim.measure.attributes(my_sim_score)

AlmaasLab/micInt documentation built on April 1, 2022, 10:37 a.m.