P281424/AMR
1
0
Fork 0
mirror of https://github.com/msberends/AMR.git synced 2025-07-08 16:42:10 +02:00
Code Releases Activity

speed improvements

Browse Source
This commit is contained in:
dr. M.S. (Matthijs) Berends
2018-07-15 22:56:41 +02:00
parent 8240959f38
commit 6eaf33baf3
13 changed files with 359 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
man/resistance.Rd
View File

@ -14,8 +14,8 @@ susceptibility(ab1, ab2 = NULL, include_I = FALSE, minimum = 30,
n_rsi(ab1, ab2 = NULL)
rsi(ab1, ab2 = NULL, interpretation = "IR", minimum = 30,
as_percent = FALSE)
rsi(ab1, ab2 = NA, interpretation = "IR", minimum = 30,
as_percent = FALSE, info = FALSE, warning = TRUE)
}
\arguments{
\item{ab, ab1, ab2}{vector of antibiotic interpretations, they will be transformed internally with \code{\link{as.rsi}}}
@ -27,6 +27,10 @@ rsi(ab1, ab2 = NULL, interpretation = "IR", minimum = 30,
\item{as_percent}{logical to indicate whether the output must be returned as percent (text), will else be a double}
\item{interpretation}{antimicrobial interpretation}
\item{info}{\emph{DEPRECATED} calculate the amount of available isolates and print it, like \code{n = 423}}
\item{warning}{\emph{DEPRECATED} show a warning when the available amount of isolates is below \code{minimum}}
}
\value{
Double or, when \code{as_percent = TRUE}, a character.
@ -37,7 +41,7 @@ These functions can be used to calculate the (co-)resistance of microbial isolat
\details{
\strong{Remember that you should filter your table to let it contain only first isolates!} Use \code{\link{first_isolate}} to determine them in your data set.
All return values are calculated using hybrid evaluation (i.e. using C++), which makes these functions 60-65 times faster than in \code{AMR} v0.2.0 and below. The \code{rsi} function is available for backwards compatibility and deprecated. It now uses the \code{resistance} and \code{susceptibility} functions internally, based on the \code{interpretation} parameter.
The functions \code{resistance}, \code{susceptibility} and \code{n_rsi} calculate using hybrid evaluation (i.e. using C++), which makes these functions 25-30 times faster than the old \code{rsi} function. This function is still available for backwards compatibility but is deprecated.
\if{html}{
\cr
To calculate the probability (\emph{p}) of susceptibility of one antibiotic, we use this formula:
@ -98,6 +102,29 @@ my_table \%>\%
genus == "Helicobacter") \%>\%
summarise(p = susceptibility(amox, metr), # amoxicillin with metronidazole
n = n_rsi(amox, metr))
# How fast is this hybrid evaluation in C++ compared to R?
# In other words: how is the speed improvement of the new `resistance` compared to old `rsi`?
library(microbenchmark)
df <- septic_patients \%>\% group_by(hospital_id, bactid) # 317 groups with sizes 1 to 167
microbenchmark(old_IR = df \%>\% summarise(p = rsi(amox, minimum = 0, interpretation = "IR")),
new_IR = df \%>\% summarise(p = resistance(amox, minimum = 0)),
old_S = df \%>\% summarise(p = rsi(amox, minimum = 0, interpretation = "S")),
new_S = df \%>\% summarise(p = susceptibility(amox, minimum = 0)),
times = 5,
unit = "s")
# Unit: seconds
# expr min lq mean median uq max neval
# old_IR 1.95600230 1.96096857 1.97981537 1.96823318 2.00645711 2.00741568 5
# new_IR 0.06872808 0.06984932 0.07162866 0.06987306 0.07050094 0.07919192 5
# old_S 1.68893579 1.69024888 1.72461867 1.69785934 1.70428796 1.84176137 5
# new_S 0.06737037 0.06838167 0.07431906 0.07745364 0.07827224 0.08011738 5
# The old function took roughly 2 seconds, the new ones take 0.07 seconds.
}
}
\keyword{antibiotics}