Finding the average of every nth row but n is not fixed.

Hello,

Consider a nx2 matrix M with

M= [

1 10

1 11

1 20

2 4

2 9

2 8

2 7

.

.

.

];

How to find the average of the second column when the first column change values, i.e. "1 rows" have 3

values while "2 rows" have 4, etc. so we need the average of the 3 first values, then the next four, etc.

Thank you.

[madhan ravi]

?? illustration needed

Answer 1

[Stephen23]

>> M = [1,10;1,11;1,20;2,4;2,9;2,8;2,8];

Method one: accumarray:

>> V = accumarray(M(:,1),M(:,2),[],@mean)
V =
  13.667
   7.25

Method two: splitapply:

>> V = splitapply(@mean,M(:,2),M(:,1))
V =
  13.667
   7.25

And if you want those values back in the matrix:

>> M(:,3) = V(M(:,1))
M =
  1  10  13.667
  1  11  13.667
  1  20  13.667
  2   4   7.25
  2   9   7.25
  2   8   7.25
  2   8   7.25

[pavlos]

@ Stephen

Thank you, it works perfectly.

But I didn`t mentioned in my question that the values of the 1st column are repeated.

Actually, the 1st column corresponds to hours in a day within a month so, 0,1,..,23 are repeated:

0 0 1. . .23 23 23 1st day

0 1 1...23 2nd day

etc.

So each day has different number of hourly values.

Pavlos

[Stephen23]

@pavlos: Then to use this method you will have to generate a unique index for each group that you want to average. You could use findgroups or discretize or the third output from unique (possibly with the 'rows' option) or something similar.