R上/下R中缺少的时间序列数据

Question

I have a set of time-series data (GPS speed data, specifically), which includes gaps of missing values where the signal was lost. For missing periods of short durations I am about to fill simply using a na.spline, however this is inappropriate with longer time periods. I would like to ramp the values from the last true value down to zero, based on predefined acceleration limits.

#create sample data frame
test <- as.data.frame(c(6,5.7,5.4,5.14,4.89,4.64,4.41,4.19,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,5,5.1,5.3,5.4,5.5))
names(test)[1] <- "speed"

#set rate of acceleration for ramp
ramp <- 6

#set sampling rate of receiver
Hz <- 1/10

So for missing data the ramp would use the previous value and the rate of acceleration to get the next data point, until speed reached zero (ie last speed [4.19] + (Hz * ramp)), yielding the following values:

3.59
2.99
2.39
1.79
1.19
0.59
0

Lastly, I need to do this in the reverse fashion, to ramp up from zero when the signal picks back up again.

Hope this is clear.

Cheers

Answer 1

It's not really elegant, but you can do it in a loop.

na.pos <- which(is.na(test$speed))

acc = FALSE
for (i in na.pos) {
    if (acc) {
        speed <- test$speed[i-1]+(Hz*ramp)
    }
    else {
        speed <- test$speed[i-1]-(Hz*ramp)
        if (round(speed,1) < 0) {
            acc <- TRUE
            speed <- test$speed[i-1]+(Hz*ramp)
        }

    }
    test[i,] <- speed
}

The result is:

   speed
1   6.00
2   5.70
3   5.40
4   5.14
5   4.89
6   4.64
7   4.41
8   4.19
9   3.59
10  2.99
11  2.39
12  1.79
13  1.19
14  0.59
15 -0.01
16  0.59
17  1.19
18  1.79
19  2.39
20  2.99
21  3.59
22  4.19
23  4.79
24  5.00
25  5.10
26  5.30
27  5.40
28  5.50

Note that '-0.01', because 0.59-(6*10) is -0.01, not 0. You can round it later, I decided not to.

Answer 2

When the question says "ramp the values from the last true value down to zero" in each run of NAs I assume that that means that any remaining NAs in the run after reaching zero are also to be replaced by zero.

Now, use rleid from data.table to create a grouping vector the same length as test$speed identifying each run in is.na(test$speed) and use ave to create sequence numbers within such groups, seqno . Then calculate the declining sequences, ramp_down by combining na.locf(test$speed) and seqno . Finally replace the NAs.

library(data.table)
library(zoo)

test_speed <- test$speed
seqno <- ave(test_speed, rleid(is.na(test_speed)), FUN = seq_along)
ramp_down <- pmax(na.locf(test_speed) - seqno * ramp * Hz, 0)
result <- ifelse(is.na(test_speed), ramp_down, test_speed)

giving:

> result     
 [1] 6.00 5.70 5.40 5.14 4.89 4.64 4.41 4.19 3.59 2.99 2.39 1.79 1.19 0.59 0.00
[16] 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5.00 5.10 5.30 5.40 5.50