In kdb+ let's say I have a table t such that
show t:([] rnk1: 1 2; rnk2: 3 4);
rnk1 rnk2
---------
1 3
2 4
I can calculate the row-wise average avg of the two columns as:
show update r:avg[(rnk1;rnk2)] from t;
show ![t;();0b;(enlist`r)!enlist (avg;(enlist;`rnk1;`rnk2))]; / alternative functional form
rnk1 rnk2 r
-----------
1 3 2
2 4 3
However, I can't seem to able to compute the row-wise exponential moving average ema using a similar approach
alpha:0.5;
show update r:ema[alpha;(rnk1;rnk2)] from t;
show ![t;();0b;(enlist`r)!enlist (ema;`alpha;(enlist;`rnk1;`rnk2))
'rank
which errors out. How do i find the row-wise ema above?
You need to flip the rnk columns in order to align them row-wise and call each ':
q)ema[0.5;1 3]
1 2f
q)ema[0.5;2 4]
2 3f
q)update r:ema'[0.5;flip(rnk1;rnk2)]from t
rnk1 rnk2 r
-------------
1 3 1 2
2 4 2 3
q)![t;();0b;enlist[`r]!enlist(ema';0.5;(flip;(enlist;`rnk1;`rnk2)))]
rnk1 rnk2 r
-------------
1 3 1 2
2 4 2 3