CODEWITHSUNDEEP
rmoving-averagearima
TCS
TCS

Reputation: 471

Calculate a moving average from circular data in r

I want to calculate a mean angle in radians (theta) from a series of angles (beta), which are stored in a dataframe called 'a':

       time beta    theta           theta2
    1   0           na              na
    2   2.426551575 na              na 
    3   5.890842799 0.678069689     0.678069689
    4   0.530641823 0.854950297     0.854950297
    5   4.794960386 -0.449975202    5.833210105
    6   1.271620156 0.104679019     0.104679019
    7   5.789548201 -0.236747291    6.046438017
    8   1.053579352 0.610520801     0.610520801
    9   0.095112206 0.218351484     0.218351484
    10  6.108843722 0.324783324     0.324783324

the betas are in radians and are generated as a random draw from the wrapped gauchy distribution. I want to calculate thetas (actually theta2, in radians). I did these by brute force, converting radians into positive and negative units with pi/-pi at 180 degrees, but was looking for a more elegant approach.

I had hoped that this would work (invoking both circular and dplyr packages):

a$theta<-mean.circular(c(a$beta,lag(a$beta),lag(a$beta,2))

But this only seems to return the last value of beta. I've tried subsetting and also within a loop (I have another field called time that is an incrementing integer)

    time    beta
    1   0
    2   2.426551575
    3   5.890842799
    4   0.530641823
    5   4.794960386
    6   1.271620156
    7   5.789548201
    8   1.053579352
    9   0.095112206
   10   6.108843722



theta<-0
bset<-c(0,0,0)
for (i in time){
  bset<-ifelse(i<3,bset,df$beta[(i-2):i])
  thetai<-mean.circular(bset)
  theta<-c(theta,ifelse(is.na(thetai),0,thetai))

theta<-theta[2:(length(time)+1)]
df<-cbind(df,theta)
}

But this also is not working. Can anyone tell me what I am doing wrong?

Upvotes: 0

Views: 795

Answers (1)

PKumar
PKumar

Reputation: 11128

I believe if I understood you correctly then you are trying to calculate moving averages of data which is circular in nature. If that is the case, You may choose raster's movingFun function to get the moving averages. Since mean is not vectorized in nature, it will always return one value.

It would be good if you paste your desired output.

Here, I choose a value of lag as 2 and circular nature as True.

library(raster)
movingFun(x, n= 2, fun = mean, circular = TRUE)

Input data:

x <- c(0,2.426551575,
       5.890842799,
       0.530641823,
       4.794960386,
       1.271620156,
       5.789548201,
       1.053579352,
       0.095112206,
       6.108843722,
       1.240271853,
       5.350922943)

Output:

> movingFun(x, n= 2, fun = mean, circular = T)
 [1] 2.6754615 1.2132758 4.1586972 3.2107423
 [5] 2.6628011 3.0332903 3.5305842 3.4215638
 [9] 0.5743458 3.1019780 3.6745578 3.2955974

Upvotes: 4

Related Questions