[R] How to take the average of multiple rows

[Marc Schwartz]

on 06/15/2008 01:18 PM ss wrote:
> Dear all,
> 
> 
> I have a matrix, called newdata1,
> 
>> dim(newdata1)
> [1] 34176    83
> 
> It looks like:
> 
>             EntrezID    Name      S1  S2   S3    S4 S5.....
> 1           4076       CAPRIN1  0.1  0.2   0.3...
> 2           139170    WDR40B  0.4  0.5   0.6...
> 3            5505    PPP1R2P1  0.3  0.3   0.7...
> 4            4076       CAPRIN1  0.7  0.3   0.2...
> 5           139170    WDR40B  null  0.8   0.4...
> 6           139170    WDR40B  null  null   0.75...
> 
> 
> If there are rows whose "EntrezID" and "Name" are exactly the same,
> I want to take the average for these rows.
> 
> There might be some 'null's in the data set. For example, there are
> three rows of the same "EntrezID" and "Name" as 139170 and WDR40B.
> For the sample called 'S1', there are three values, 0.4, null, null. For
> this
> scenario, I want to keep the final value as 0.4 For the sample 'S2', the
> values
> are 0.5, 0.8, null. For this, I want to ignore 'null', and take the average
> of
> 0.5 and 0.8, the final value is (0.5+0.8)/2=0.65. For the sample 'S3', there
> is no 'null', so just take the average as (0.6+0.4+0.75)/3=0.5833.
> 
> Can you show me how to do this?
> 
> I appreciate!

If your data file is exactly the way you have it above, you first want 
to convert the 'null' entries to NA so that they are treated as missing 
values by R.

Thus:

DF <- read.table(YourFileName, header = TRUE, na.strings = "null")

 > DF
   EntrezID     Name  S1  S2   S3
1     4076  CAPRIN1 0.1 0.2 0.30
2   139170   WDR40B 0.4 0.5 0.60
3     5505 PPP1R2P1 0.3 0.3 0.70
4     4076  CAPRIN1 0.7 0.3 0.20
5   139170   WDR40B  NA 0.8 0.40
6   139170   WDR40B  NA  NA 0.75


Then use aggregate():

 > aggregate(DF[, -c(1:2)], by = list(DF$EntrezID, DF$Name),
             mean, na.rm = TRUE)
   Group.1  Group.2  S1   S2        S3
1    4076  CAPRIN1 0.4 0.25 0.2500000
2    5505 PPP1R2P1 0.3 0.30 0.7000000
3  139170   WDR40B 0.4 0.65 0.5833333


See ?read.table, ?aggregate and ?mean for more information. Take note of 
the 'na.rm' argument in ?mean.

HTH,

Marc Schwartz