## Average columns based on header name

# 1
03-17-2016
Average columns based on header name

Hi Friends,

I have files with columns like this. This sample input below is partial.

Please check below for main file link. Each file will have only two rows.

Basically, here is what I need to be done.

a. Start from second column which is 0.4% here.
b. Go until you hit "10" in the header name. If the header name is exactly 10.0%, then include that column too. If not, only include until the column before it. In this example, since we have 10.1% (29th column), we will be including columns starting from 0.4%(second) until 9.8% which is the 28th column. If the 29th column was to be 10.0%, then it would have been included too.
c. Average the values for these respective columns in the second row (data is not presented here - please click this link for total dataset - https://goo.gl/W8jND7). In this example, starting from 0.4%(second column) till 9.8%(28th column).
d. In the output, print first column which is "Gene", and this average value with column header being

e. Then start from 10.1% (29th column) and check until you hit "20" in the header name. Repeat steps b through d. And print output as

Repeat this until you have

f. After you hit 100%, it means one dataset is done.

g. If you observe my column header carefully here, there is another 0.4%-100% columns after the first 100%. I will be having 13 of these 0.4%-100%s in the input file at the above link.

i. I have multiple files, the headers can be

It varies from file to file. But the logic of averaging(if you hit "10", "20", etc) is always the same. And the number of samples 13 is also same which means each file will have 100%s for 13 times.

P.S: A Bonus of 1000 bits will be awarded to the effectively working solution.
 jacobs.smith View Public Profile for jacobs.smith Find all posts by jacobs.smith
# 2
03-17-2016
I do not understand what output you are trying to produce.

For each input gene line do you want:
1. one output line with the gene number and 10 averages from all 13 samples,
2. thirteen output lines with the gene number and the 10 averages from one sample on each line, or
3. one output line with the gene number and 130 averages where each set of 10 averages comes from one sample?
Can you show us the exact output you're hoping to produce from the data provided in your sample input for genes 1 and 2?

Was the data for gene 3 in your sample truncated, or will some inputs have missing fields that should be treated as zero values when computing the averages?
This User Gave Thanks to Don Cragun For This Post:
jacobs.smith (03-17-2016)
 Don Cragun View Public Profile for Don Cragun Find all posts by Don Cragun
# 3
03-17-2016
Quote:
Originally Posted by Don Cragun
I do not understand what output you are trying to produce.

For each input gene line do you want:
1. one output line with the gene number and 10 averages from all 13 samples,
2. thirteen output lines with the gene number and the 10 averages from one sample on each line, or
3. one output line with the gene number and 130 averages where each set of 10 averages comes from one sample?
Can you show us the exact output you're hoping to produce from the data provided in your sample input for genes 1 and 2?

Was the data for gene 3 in your sample truncated, or will some inputs have missing fields that should be treated as zero values when computing the averages?
Hi Don,

Thank you for your response. Good questions and I am glad at least you replied. Thank you so much.

I want "one output line with the gene number and 130 averages where each set of 10 averages comes from one sample"

Since the data is very big, I will use the below small example assuming I have only one sample.

Here the 2.6 is coming from averaging 1 in second column through 1 in 11th column. 5.9 is coming from averaging from 22.5(12th column) till 7 (21st column). Please remember that if the 11th column was to be any value greater than 10.0% like 10.1% or 10.8% or anything, then we will be averaging only until the 10th column.

The main input file from the link that I gave in my earlier post, has 13 samples' 1% to 100% values and only two rows. In the output file, we will have Gene column plus 130 average values. Makes sense?

There will be no gene3 in an input file. It will always be 2 rows (first line is header and second line is the values for averaging) with 13 sample being chopped into different varying percentages. There will be no missing fields. For every column header, there is a value associated with it in the input file.

Also I have a batch of files like 4000. I give a folder with the extension, and the script should be reading each file and following the averaging conditions of differentiating exact 10.0% (to consider for average) and any other values greater than 10.0% like 10.1% or 10.2% or 10.3% etc (to not consider this column and the column before it).

Coming to what I have tried so far, I have been trying to read the headers and print each set into a different file and then do the computation and put it back and then move to the other sample. This seems to be very time taking.

All your time and understanding is highly appreciated.

 jacobs.smith View Public Profile for jacobs.smith Find all posts by jacobs.smith
# 4
03-17-2016
This is a hell of a problem, by sheer data volume. I'm brain dead now; can't verify the results are correct although I think/hope they are. Try Please do some validation and come back with its result.
This User Gave Thanks to RudiC For This Post:
jacobs.smith (03-17-2016)
 RudiC View Public Profile for RudiC Find all posts by RudiC
# 5
03-17-2016
Quote:
Originally Posted by RudiC
This is a hell of a problem, by sheer data volume. I'm brain dead now; can't verify the results are correct although I think/hope they are. Try Please do some validation and come back with its result.
Rudic,

You are one piece of a beast! The solution is rocking!

Two quick requests.

1. You are printing the sum and not averages. I just checked and your script is working fine. I am scared to touch this script. Can I print SUM/NR to get the average?

2. Can you make this oneliner? Please...please...please. Again, I can make it. But I don't want to F*** this up. A BIG THANKS MY FRIEND.

When Don asked me how to print the output file, I asked him to print me 130 columns, but your solution seems to be perfectly suitable to my further downstream analysis.

Thank you so much.
 jacobs.smith View Public Profile for jacobs.smith Find all posts by jacobs.smith
# 6
03-17-2016
As I said, I'm braindead. I missed the average, sorry. Try printing SUM/(L[i+1]-L[i]). That's not OK for the very first entry as the first field is summed, too, as its value is zero, so the element count is too high. But - I'm scared to touch it, either.

No one liner. Whereever a oneliner can be accommodated, a multiliner will fit, too. Use a function. Source a file.

---------- Post updated at 22:40 ---------- Previous update was at 22:28 ----------

Try this one - the field 1 error should be removed. Please perform a careful validation of the results and come back: ---------- Post updated at 23:39 ---------- Previous update was at 22:40 ----------

Actually, transposing the file like simplified the logics seriously. Piping the paste result into gives you an idea - results may still need a bit polishing.
This User Gave Thanks to RudiC For This Post:
jacobs.smith (03-17-2016)
 RudiC View Public Profile for RudiC Find all posts by RudiC
# 7
03-17-2016
Hi Rudic,

For some of the files, I am seeing this error with the first long code

Any thoughts?

But the paste tweak is perfect!

Last edited by jacobs.smith; 03-17-2016 at 09:59 PM..
 jacobs.smith View Public Profile for jacobs.smith Find all posts by jacobs.smith