Description Usage Arguments Details Value Author(s) References Examples

Transforms a weighted matrix into another one constrained into a common scale of relative strength. Values are reweighted by a combination of rewards and punishments. For a given link, its relative strength increases with the number of triads where it is the strongest and decreases with the number of triads where it is the weakest.

1 2 |

`wm` |
Square and symmetric matrix filled with valued score of spatial affinity. |

`similarity` |
Logical, depending on the existence of direct or inverse relationship between matrix scores and the spatial affinity between species (i.e. TRUE: high scores - high affinity; FALSE: low scores - high affinity). |

`t1t2` |
Non-negative numeric vector, conventionally of length 2. |

`normalized` |
Logical, if TRUE output values are linearly transformed into the scale [0, 1]. |

The vector `t1t2`

holds the thresholds to delimit intervals for meaningful
association. If `similarity`

is TRUE, any value of the original
matrix lower than min(t1t2) is discarded for further consideration because it
represents the lower bound for meaningful association. On the other hand, any value
of the input matrix equal or higher than the upper bound (i.e. max(t1t2)) is deemed
to be maximally similar and coded 1 in the output. If `similarity`

is FALSE,
any link lower than min(t1t2) represent close affinity between species, and conversely
any link equal or higher than max(t1t2) denote no affinity at all between its
endvertices.

The dichotomization is a common procedure to transform weighted matrices into binary ones.
Values are set 1 or 0 depending on the location of some cutoff value. This function
also enables us to perform a hard thresholding when the parameter `t1t2`

has
been fed with a single value. Under this setting, the final reweighted matrix will
be binary. If `similarity`

is TRUE (FALSE), entries are 1 if original values are equal
or higher (lower) than `t1t2`

, otherwise entries are 0.

A similarity matrix composed of reweighted values. Values are constrained into the interval [0, 1] under the option of normalization. Here, null association is coded with a zero (0), whereas the unity (1) corresponds to the optimal meaningful association.

Daniel A. Dos Santos <dadossantos@csnat.unt.edu.ar>

Dos Santos D.A., Cuezzo M.G., Reynaga M.C., Dominguez E. 2011. *Towards a Dynamic
Analysis of Weighted Networks in Biogeography.* Systematic Biology (in press).

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 | ```
#Example to show an interesting numerical property of the reweighting formula, i.e.
#its convergence to the same rescaled matrix after several self-iterations
nitems <- 25
ej <- matrix(0, nitems, nitems)
ej[row(ej) > col(ej)] <- runif(nitems*(nitems - 1)/2)
t(ej) + ej -> ej
diag(ej) <- 1
### Display three graphics
op <- par(mar = rep(3, 4), mfrow = c(3, 1))
plot(unlist(ej),unlist(reweight(ej)), xlab = "Input matrix", ylab = "Reweighted matrix", main = "First reweighting")
#Iterative reweighting
histcor <- c()
#Perform 100 iterations
for(i in 1:100) {
reweight(ej) -> rej
b <- c()
for(k in 1:(nrow(ej) - 1))
for(j in (k + 1):nrow(ej)) {
a <- (ej[k, j] >= ej[k, ]) + (rej[k, j] >= rej[k, ])
b <- c(b, sum(a==2)/sum(a > 0))
}
histcor <- c(histcor, mean(b))
ej <- rej
}
plot(unlist(ej),unlist(reweight(ej)), xlab = "Input matrix", ylab = "Reweighted matrix",
main = "Reweighting after several iterations")
plot(histcor, xlab = "Iteration", ylab = "Cross adjustment",
main = "Evolution of the resemblance between input and reweigthed network")
## At end of plotting, reset to previous settings:
par(op)
``` |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.