Fixed NA issue introduced by get_together_sticky that was breaking tests. NA handling is now a bit simplified. If any individuals have NAs they are considered 'not together' with anyone, and any ambiguous times that are not connected to a period of connectedness are considered 'apart'. Need to test a few more edge cases involving NAs and periods of together vs apartness at the beginning/end of sequences.

Author: Ariana Strandburg-Peshkin

R/get_together_sticky.R

@@ -95,7 +95,7 @@ get_together_sticky <- function(together_inner, together_outer){
   }
 
   #any NAs are not considered together
-  together_ij[which(is.na(together_inner))] <- NA
+  together_ij[which(is.na(together_ij))] <- FALSE
 
   return(together_ij)
 }

R/identify_splits_and_merges.R

@@ -13,9 +13,7 @@
 #'
 #'NA handling:
 #' Individuals are considered not together if either of their positions is not known. If a period of connectedness runs into an NA,
-#'individuals will be considered as connected up until that NA. However, if an
-#'ambiguous period (between the two thresholds) runs into an NA, individuals
-#'will be considered
+#'individuals will be considered as connected up until that NA.
 #'Once connectedness of dyads is determined, merge dyads together into groups by
 #'using DBSCAN on 1 - `together` as the distance matrix, with `eps` equal to
 #'something small (.01 in the code).

inst/tinytest/test-known-issues.R

@@ -1,5 +1,5 @@
 # skip
-#exit_file()
+exit_file()
 
 # Analyse split and merge
 # before time is not tidx - time windwo