Alexander Alexander - 3 months ago 8
R Question

Finding Row number of Consecutive decreasing values in data

I need to detect from data the first element of the first sequence of length 5 of consecutively decreasing numbers. There is a similar post here but when I applied to my data it failed.

set.seed(201)
az <- c(sort(runif(10,0,0.9),decreasing = T),sort(runif(3,-0.3,0),decreasing = T),sort(runif(3,-0.3,0),decreasing = F),sort(runif(4,-0.3,0),decreasing = T),sort(runif(4,-0.3,0),decreasing = F),sort(runif(6,-0.3,0),decreasing = T))
tz <- seq(1,length(az))
df <- data.frame(tz,az=round(az,2))


enter image description here

In the figure above it would be somewhere around tz = 25.

The post says that this function need to improve and so far I cannot get my desired result!

getFirstBefore<-function(x,len){
r<-rle(sign(diff(x)))
n<-which(r$lengths>=len & r$values<0)
if(length(n)==0)
return(-1)
1+sum(r$lengths[seq_len(n[1]-1)])
}

df1 <- df%>%
mutate(cns_tz=getFirstBefore(az,5))

tz az cns_tz
#1 1 0.56 4
#2 2 0.55 4
#3 3 0.33 4
#4 4 0.33 4
#5 5 0.26 4
#6 6 0.15 4
#7 7 0.12 4
#8 8 0.09 4
#9 9 0.04 4
#10 10 0.04 4
#11 11 -0.10 4
#12 12 -0.12 4
#13 13 -0.16 4
#14 14 -0.16 4
#15 15 -0.14 4
#16 16 -0.14 4
#17 17 -0.13 4
#18 18 -0.15 4
#19 19 -0.22 4
#20 20 -0.30 4
#21 21 -0.12 4
#22 22 -0.12 4
#23 23 -0.11 4
#24 24 -0.07 4
#25 25 -0.05 4
#26 26 -0.09 4
#27 27 -0.10 4
#28 28 -0.15 4
#29 29 -0.17 4
#30 30 -0.22 4

Answer

We can use rleid from data.table

library(data.table)
n <- 5
v1 <- setDT(df)[sign(az)<0,  .I[which(.N==n)] , rleid(c(1, sign(diff(az))))]$V1[1L]
v1
#[1] 26
df[, cnz_tz := v1]

Or another option is shift with Reduce

setDT(df)[, cnz_tz := .I[Reduce(`&`, shift((az - shift(az, fill=az[1])) < 0,
                            0:4, type = "lead", fill=FALSE)) & sign(az) < 0][1]]

We can also use rleid in dplyr

library(dplyr)
v1 <- df %>% 
         group_by(rl= rleid(c(1, sign(diff(az))))) %>% 
         mutate(rn =  sign(az) < 0 & n()==5) %>% 
        .$rn %>%
         which() %>% 
         head(., 1)
v1
#[1] 26
df %>%
   mutate(cnz_tz = v1)