long_to_wide() turns a long event table (one row per action) into wide
sequences, with one row per sequence and one column per position
(V1, V2, ...). Each sequence's actions are ordered by time_col, and
the k-th action fills column k.
Usage
long_to_wide(
data,
id_col = "Actor",
time_col = "Time",
action_col = "Action",
time_prefix = "V",
fill_na = TRUE
)Arguments
- data
A data frame in long format, one row per action.
- id_col
The column identifying each sequence. Default
"Actor"; a missing column raises an error.- time_col
The column that orders the actions within a sequence. Default
"Time". Its values only order the actions; they do not set the column numbers, so gaps between time points are not kept. Actions with the same time keep their row order. When the column is absent, the row order is used.- action_col
The column holding the actions (states). Default
"Action"; a missing column raises an error.- time_prefix
The prefix of the position columns in the output. Default
"V".- fill_na
How to handle sequences of different lengths. With
TRUE(default), shorter sequences are padded withNA, so every row has the columns of the longest sequence. WithFALSE, only the positions present in every sequence are kept, so longer sequences are cut to the length of the shortest.
Value
A data frame in wide format, one row per sequence in the sorted
order of id_col: the id_col column followed by the position columns
(V1, V2, ..., named with time_prefix) holding the action at each
position. Other columns of data are dropped. A missing action stays a
missing cell.
Details
The wide format is the one build_network(), tna::tna() and related
functions read without column declarations, since the position columns
form a numbered run; build_network() sets the id_col column aside as
metadata, with a message.
See also
wide_to_long for the reverse conversion,
prepare_for_tna for preparing data for TNA analysis.