How can I do the rolling sum of 5 elements for each column? I have a dataset like this, and would be greatful, someone can help me to write a code for rolling sum of 5 elements of each column.

da =

         39.00         34.00         40.00         38.00        151.00
         47.00         33.00         35.00         48.00        163.00
         42.00         46.00         43.00         51.00        182.00
         40.00         30.00         37.00         39.00        146.00
         39.00         30.00         42.00         30.00        141.00
         40.00         29.00         45.00         35.00        149.00
         25.00         35.00         47.00         30.00        137.00
         36.00         28.00         37.00         28.00        129.00
         24.00         37.00         37.00         34.00        132.00

Answer 1

[Stephen23]

Use conv2:

>> X = conv2(da,ones(5,1))

Note that there are several options for selecting what to do with the sums at the edges: you need to read the documentation carefully and pick the one that suits your needs best. For example the option 'valid' only keeps the values for which five values are summed, thus excluding the edges values:

>> X = conv2(da,ones(5,1),'valid')
X =
   207   173   197   206   783
   208   168   202   203   781
   186   170   214   185   755
   180   152   208   162   702
   164   159   208   157   688

If you want the sums starting from the first row but keeping only as many rows as da has, then try using the option 'full':

>> X = conv2(da,ones(5,1),'full');
>> X(1:size(da,1),:)
ans =
    39    34    40    38   151
    86    67    75    86   314
   128   113   118   137   496
   168   143   155   176   642
   207   173   197   206   783
   208   168   202   203   781
   186   170   214   185   755
   180   152   208   162   702
   164   159   208   157   688

[Nadeera Gunartna]

Thank you very much Stephen

[Nadeera Gunartna]

when i run my code it display the answer and after the answer it shows me "Index exceeds matrix dimensions" and then rest of the code is not running. can you give me some advices to avoid this error?

[Stephen23]

Probably you are calling conv2 with the option 'valid', in which case the output matrix is smaller than da, and so you cannot use this code afterwards:

X(1:size(da,1),:) % if X is smaller than da this will be an error!

Note my example using 'valid' does not use this line of code :)

I just gave that line as an example: it is not strictly required for you to use it! It all depends on what values you need and what size matrix you need. Because you have not told us exactly these specifications I showed you several options.

[Nadeera Gunartna]

yeah.. I used 'valid' , in my case I have to use it because I need to plot graphs of every column. if I used 'valid', is there no way to avoid that error? if there is any command please let me know. Thanks

[Stephen23]

Obviously if you try to access an element of a matrix that does not exist then this is an error.

The solution depends on you. You have not given enough information for me to know what you need, or what exactly you are plotting. If you actually show us your code then we know what you are doing.

The different conv2 options give different size output matrices. I do not know what size you want because you have not told us what size you want the output to be.

I am going to make a guess that you want the output to be the same size as the input da. These will work, but keep in mind they give different results and only you know what results you need. Whatever results you need don't forget check them by hand!

 X =conv2(da,ones(5,1),'same') % X same size as da

or

 X = conv2(da,ones(5,1),'full');
 X = X(1:size(da,1),:) % shortened X to same size as da

Possibly you want to shorten da to be the same length as the valid output:

 X = conv2(da,ones(5,1),'valid');
 db = da(1:size(X,1),:) % shortened da to same size as X

[Nadeera Gunartna]

it works. thanks