full_mm: Create a 3-dimensional mixingm matrix
In mbojan/isnar: Introduction to Social Network Analysis with R

Description Usage Arguments Value Examples

Given contact layer of the mixing matrix compute a full 3-dimensional mixing matrix by rebuilding the non-contact layer from supplied group sizes and other arguments.

1	full_mm(cl, gsizes, directed = TRUE, loops = FALSE)

`cl`	numeric matrix with contact layer of the mixing matrix
`gsizes`	numeric vector or matrix with group sizes, see Details.
`directed`	logical, whether the network is directed
`loops`	logical, whether loops (self-ties) are allowed Contact layer of the mixing matrix is a cross-classification of ties according to the attributes of tie sender (ego) and tie receiver (alter). Classically, the same attribute is used for ego and alter resulting in a square mixing matrix. In such cases `cl` shoud be square with `cl[i,j]` being the number of ties from actors in group `i` to actors in group `j` and `gsizes` should be a numeric vector with number of nodes in each group. Consequently, it must hold that `length(gsizes) == nrow(cl) == ncol(cl)`. In more general case we can use different node attributes for ego and alter. Then `cl` does not have to be square. In such cases `gsizes` should be a cross-tabulation of nodes according to their values on both attributes. See Examples.

full_mm returns a full three-dimenstional mixing matrix as an array with dim attribute equal to c( nrow(cl), ncol(cl), 2 ).

### Square example

# Contact layer of the mixing matrix
mm1 <- matrix( c( 20, 10, 5,
                 12, 30, 10,
                 3, 11, 25 ),
              byrow=TRUE, ncol=3, nrow=3)
dimnames(mm1) <- list(ego=letters[1:3], alter=letters[1:3])
mm1

# Assuming some group sizes
gs1 <- c(a=9, b=12, c=10)

# Full mixing matrix
full_mm( mm1, gs1)




### Non-square example

# Mixing matrix
# Now using different attributes for ego and alter
mm2 <- cbind(mm1, c(20, 10, 5))
colnames(mm2) <- LETTERS[1:4]
names(dimnames(mm2)) <- c("ego", "alter")
mm2

# Create artificial distribution of attributes
set.seed(123)
a1 <- sample(letters[1:3], sum(gs1), replace=TRUE, prob=gs1/sum(gs1))
table(a1)
a2 <- sample(LETTERS[1:4], sum(gs1), replace=TRUE)
table(a2)
(x <- table(a1, a2))         # Cross-tablulation

# Full mixing matrix
full_mm( mm2, gsizes=x)