Using TTR package to calculate Exponential moving average

ganesh reddy picture ganesh reddy · Sep 24, 2012 · Viewed 15.2k times · Source

Curious if anyone used it. I did a simple EMA operation on a time series. But wasn't able to reconcile very well.

I read that the value of the update constant = 2/(N+1). I defined x = 1:20, and did EMA(x,5). Then I did an EMA computation using the recursive computation. The two results don't really line up

The function returns

EMA(x,5)
 [1] NA NA NA NA  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18

And my little thing gives me,

EMA
 [1]  1.000000  1.333333  1.888889  2.592593  3.395062  4.263374  5.175583  6.117055  7.078037  8.052025  9.034683 10.023122 11.015415 12.010276 13.006851 14.004567
[17] 15.003045 16.002030 17.001353 18.000902

Answer

Sameer picture Sameer · Sep 24, 2012

To get the answer you are looking for you would need to write

TTR::EMA(1:20, n=1,  ratio=2/(5+1))

 [1]  1.000000  1.333333  1.888889  2.592593  3.395062  4.263374  5.175583
 [8]  6.117055  7.078037  8.052025  9.034683 10.023122 11.015415 12.010276
[15] 13.006851 14.004567 15.003045 16.002030 17.001353 18.000902

If you call TTR::EMA(1:20, n=5) it is equivalent to calling

TTR::EMA(1:20, n=5, ratio=2/(5+1))

This will put NA's in the first 4 places and then the 5th place will be the simple mean of the first 5 entries. (i.e. 3 in this case). Then the EMA algorithm will start. The 6th place will be 6 * 2 / 6 + 3 * 4 / 6 = 4. The 7th place will be 7 * 2 / 6 + 4 * 4 / 6 = 5. Etc...

You can see the exact algorithm here