Convert netify objects back to dyadic data frames

unnetify (also available as netify_to_df) reverses the netify transformation by converting network objects back into dyadic (edge-level) data frames. This function combines network structure with any associated nodal and dyadic attributes, creating a data frame where each row represents a dyad (edge) with all relevant attributes attached.

Usage

unnetify(netlet, remove_zeros = FALSE)

netify_to_df(netlet, remove_zeros = FALSE)

Arguments

netlet: A netify object to be converted to dyadic format.
remove_zeros: Logical. If TRUE, removes dyads with zero edge weights from the output, resulting in a data frame of only non-zero relationships. If FALSE (default), includes all possible dyads in the network.

Value

A data frame with one row per dyad containing:

from: Name/ID of the first actor in the dyad
to: Name/ID of the second actor in the dyad
time: Time period (for longitudinal networks)
weight column: Edge weight values (column named after the weight parameter used in netify)
from_id: Unique identifier combining actor and time (longitudinal only)
to_id: Unique identifier combining actor and time (longitudinal only)
Dyadic attributes: Any edge-level covariates from the original data
Nodal attributes: Actor-level covariates merged onto dyads:
- For directed networks: suffixed with "_from" and "_to"
- For undirected networks: suffixed with "_dyad"

Details

This function essentially reverses the netify process, making it useful for:

Exporting network data for analysis in other software
Creating dyadic datasets for regression analysis
Inspecting network data in familiar data frame format
Merging network results back with other dyadic covariates

For directed networks, nodal attributes are attached twice - once for the sender (suffixed "_from") and once for the receiver (suffixed "_to"). This allows for modeling sender and receiver effects separately.

For undirected networks, nodal attributes are attached once per dyad with the suffix "_dyad", since there is no meaningful distinction between sender and receiver.

The function handles both cross-sectional and longitudinal netify objects, automatically detecting the structure and adjusting the output accordingly.

Note

For large networks, setting remove_zeros = FALSE can result in very large data frames, as it includes all possible dyads (n × (n-1) for directed networks or n × (n-1) / 2 for undirected networks).

Author

Cassy Dorff, Shahryar Minhas

Examples

# Load example data
data(icews)

# Create a netify object with attributes
icews_10 <- icews[icews$year == 2010,]

verbCoop_net <- netify(
  icews_10,
  actor1 = 'i', actor2 = 'j',
  symmetric = FALSE,
  weight = 'verbCoop',
  nodal_vars = c('i_polity2', 'i_log_gdp'),
  dyad_vars = c('verbConf', 'matlConf')
)

# Convert back to dyadic data frame
dyad_df <- unnetify(verbCoop_net)

# Examine structure
head(dyad_df)
#>        from          to time verbCoop verbConf matlConf     from_id
#> 1   Albania Afghanistan    1        5        0        1   Albania_1
#> 2   Algeria Afghanistan    1        5        1        0   Algeria_1
#> 3    Angola Afghanistan    1        0        0        0    Angola_1
#> 4 Argentina Afghanistan    1        2        0        0 Argentina_1
#> 5   Armenia Afghanistan    1       10        0        0   Armenia_1
#> 6 Australia Afghanistan    1      200       12       40 Australia_1
#>           to_id i_polity2_from i_polity2_to i_log_gdp_from i_log_gdp_to
#> 1 Afghanistan_1              9           NA       23.06701     23.49884
#> 2 Afghanistan_1              2           NA       25.67186     23.49884
#> 3 Afghanistan_1             -2           NA       24.97089     23.49884
#> 4 Afghanistan_1              8           NA       27.03815     23.49884
#> 5 Afghanistan_1              5           NA       22.86492     23.49884
#> 6 Afghanistan_1             10           NA       27.79745     23.49884
names(dyad_df)
#>  [1] "from"           "to"             "time"           "verbCoop"      
#>  [5] "verbConf"       "matlConf"       "from_id"        "to_id"         
#>  [9] "i_polity2_from" "i_polity2_to"   "i_log_gdp_from" "i_log_gdp_to"  

# Remove zero-weight dyads for more compact output
dyad_df_nonzero <- unnetify(verbCoop_net, remove_zeros = TRUE)
nrow(dyad_df_nonzero)  # Much smaller than full dyadic dataset
#> [1] 9976

# Note how nodal attributes are added
# For directed network: _from and _to suffixes
head(dyad_df[, c("from", "to", "i_polity2_from", "i_polity2_to")])
#>        from          to i_polity2_from i_polity2_to
#> 1   Albania Afghanistan              9           NA
#> 2   Algeria Afghanistan              2           NA
#> 3    Angola Afghanistan             -2           NA
#> 4 Argentina Afghanistan              8           NA
#> 5   Armenia Afghanistan              5           NA
#> 6 Australia Afghanistan             10           NA

# Longitudinal example
verbCoop_longit <- netify(
  icews,
  actor1 = 'i', actor2 = 'j', 
  time = 'year',
  symmetric = FALSE,
  weight = 'verbCoop',
  nodal_vars = c('i_polity2', 'i_log_gdp')
)

# Convert longitudinal network
dyad_df_longit <- unnetify(verbCoop_longit, remove_zeros = TRUE)

# Check time periods are included
table(dyad_df_longit$time)
#> 
#> 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 
#> 8692 8887 9514 9346 9429 9739 9988 9854 9976 9708 9841 9911 9774 

# Each dyad now has associated time period
# Note: weight column is named after the weight variable (verbCoop)
head(dyad_df_longit[, c("from", "to", "time", "verbCoop")])
#>        from          to time verbCoop
#> 1   Albania Afghanistan 2002        8
#> 2    Angola Afghanistan 2002        1
#> 3 Argentina Afghanistan 2002        2
#> 4   Armenia Afghanistan 2002        3
#> 5 Australia Afghanistan 2002      127
#> 6   Austria Afghanistan 2002       21

# Use the output for further analysis
if (FALSE) { # \dontrun{
# For example, regression analysis
lm(verbCoop ~ i_polity2_from + i_polity2_to + verbConf, data = dyad_df)
} # }