Creating network objects from adjacency lists

[mbojan]

PLEASE also add a NODELIST (ego, alter1, alter2). Empirical survey data mostly comes as a NODELIST. I have spent so much time with getting nodelists into statnet:

Originally posted by @jdohmen in #64 (comment)

[mbojan]

A primitive tidyversian prototype seems straightforward:

txt <- "ego alter1 alter2 alter3
1 2 3 4
2 1 3
3 2 1
4 1"

d <- read.table(textConnection(txt), as.is = TRUE, fill = TRUE, header = TRUE)
d
#>   ego alter1 alter2 alter3
#> 1   1      2      3      4
#> 2   2      1      3     NA
#> 3   3      2      1     NA
#> 4   4      1     NA     NA

stopifnot({
  library("dplyr")
  requireNamespace("tidyr")
})
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
#> Loading required namespace: tidyr

d |>
  tidyr::pivot_longer(- ego, values_to = "to") |>
  filter(!is.na(to)) |>
  select(from = ego, to, nomination = name)
#> # A tibble: 8 × 3
#>    from    to nomination
#>   <int> <int> <chr>     
#> 1     1     2 alter1    
#> 2     1     3 alter2    
#> 3     1     4 alter3    
#> 4     2     1 alter1    
#> 5     2     3 alter2    
#> 6     3     2 alter1    
#> 7     3     1 alter2    
#> 8     4     1 alter1

... which gives an edgelist ready to be consumed by as.network() data frame method.

The input in the example assumes that adjacency information is "reciprocated"/, i.e. if 1 and 2 are connected then 2 will appear adjacent in the row for 1 and vice versa. Thus each tie unnecessarily appears twice. I believe this corresponds to the adjacency list representation called-upon in some graph algorithms.

On the other hand, in case of "survey data" (presumably a name generator in an egocentric study) ties should be interpreted as directed (thus not reciprocated by design).