Counting the number of times a value change signs with R

Question

I am very new with R but encouraged with it because I find it accessible although I am not a programmer. I am trying to get around the following issue: I need to count how many times a value change signs in a column and then sort the results by Path (The example for a table is below- Path is a factor). I can figure how to sort the data once I eventually get them but have yet to figure out the count for the number of times a + sign becomes - and a - sign becomes a + one. Any suggestion?

Test <- structure(list(Path = c(1L, 1L, 1L, 2L, 2L, 2L), Direction = c(-3.84089, 
-1.12258, 1.47411, -1.47329, 5.4525, 10.161)), .Names = c("Path", 
"Direction"), class = "data.frame", row.names = c(NA, -6L))
head(Test)
#>   Path    Direction
#> 1    1     -3.84089
#> 2    1     -1.12258
#> 3    1      1.47411
#> 4    2     -1.47329
#> 5    2      5.4525
#> 6    2     10.161

Answer 1

I think what you are looking for is

 sum(diff(sign(X)) != 0)

where X is the vector, in your case, dat$Direction which you're trying to count sign changes.

---

If you want to calculate by Path, you can use the by function, or convert your data.frame to a data.table and use the built in by capabilities.

Example:

assuming X is your original data.frame

# I'm adding another row to the data, just to show that it works 
#    (ie, giving the two Path values a different number of rows)
X <- rbind(X, c(2, -5))

# convert to a data.table
library(data.table)
DT <- data.table(X)

# count the number of changes, per path
DT[, changes := sum(diff(sign(Direction)) != 0), by=Path]

Edit (re comment about factors):

if Direction is a factor, it will need to be converted to numeric first. You can do this using

DT[, Direction := as.numeric(Direction)]

Results:

DT

       Path Direction changes
  1:    1  -3.84089       1
  2:    1  -1.12258       1
  3:    1   1.47411       1
  4:    2  -1.47329       2
  5:    2   5.45250       2
  6:    2  10.16100       2
  7:    2  -5.00000       2