How to apply arrayfun for multiple column

Dear all, I am using following code for the one column in a data matrix. however, I have 60 columns and 6892 rows in the data matrix. How to apply the collowing code on the all columns, please suggest.

n = 6892; % total number of rows
m = 100; % window size
h = arrayfun(@(k)genhurst(y(k:m+k-1)),1:30:(n-m+1),'UniformOutput',false); % 30 is the forward shifting size
HH=h.'; 

Answer 1

[Guillaume]

If you really want to use arrayfun you could do it like this:

windowsize = 100;  %why call it m when window size is a lot clearer?
[rows, cols] = ndgrid(1:30:size(y, 1)-windowsize+1, 1:size(y, 2));
HH = arrayfun(@(row, col) genhurst(y(row:row+windowsize-1, col)), rows, cols, 'UniformOutput', false);

or you could just use an explicit loop which might be clearer and faster.

[BISWA BHUYAN]

Woww...this is very nice. But, please give an example of explicity loop. I do not know what is this?

[Guillaume]

windowsize = 100;
startrows = 1:30:size(y, 1);
HH = cell(numel(startrows), size(y, 2));
for col = 1:size(y, 2)
    for row = 1:numel(startrows)
        HH{row, col} = genhurst(y(startrows(row):startrows(row)+windowsize-1, col));
    end
end

[BISWA BHUYAN]

Thank you very much. I have one more similar problem, but with a different manner. I have atatched my data set. In the data set six variables are there (from x1 to x6) and one date column. The date ranges from 03-01-1994 to 29-12-1995. It is in date-month-year format.

Now i want to use function 'genhurst' on the each month in each year. This will be used for all variables.

To be more specific, in the first step data will be split on the basis of month and year. So, we will get 12 months in a year. In the second step, 'genhurst' will be used for each month and year. In the third step, this will be performed for all variables. If my question is not clear, please ask me, I will explain.

Please suggest.

[Guillaume]

The easiest would be to import your data as a table or timetable, then use groupsummary or retime:

data = readtable('data.xlsx');
monthly_data = groupsummary(data, 'Date', 'month', @genhurst)

or

data = readtimetable('data.xlsx');
monthly_data = retime(data, 'monthly', @genhurst)

[BISWA BHUYAN]

Query on above suggestion because i got error in the above suggestion. When i used your suggestion,

>> windowsize = 100;

startrows = 1:30:size(y, 1);

HH = cell(numel(startrows), size(y, 2));

for col = 1:size(y, 2)

for row = 1:numel(startrows)

HH{row, col} = genhurst(y(startrows(row):startrows(row)+windowsize-1, col));

end

end

I got the following error.

Index exceeds matrix dimensions.

However, when i used below code, i am not getting any error.

windowsize = 100;

[rows, cols] = ndgrid(1:30:size(y, 1)-windowsize+1, 1:size(y, 2));

HH = arrayfun(@(row, col) genhurst(y(row:row+windowsize-1, col)), rows, cols, 'UniformOutput', false);

The above code is performing without date column. However, when there is 'm' number of date column with 'm' number of variable available, then the above code is not working. For example, the atatched excel file contains two date column and two variables column. When the following code applies, I am getting errors.

data = readtable('data.xlsx');

windowsize = 100;

startrows = 1:30:size(date, 1);

HH = cell(numel(startrows), size(date, 1));

for col = 1:size(date, 1)

for row = 1:numel(startrows)

HH{row, col} = genhurst(date(startrows(row):startrows(row)+windowsize-1, col));

end

end

Error using date

Too many input arguments.

In the above code i want to store the outcome of first window on the date row of 'windowsize+1' along with the respective date in a new matrix.

Please suggest

[Guillaume]

I made a silly mistake in the explicit loop code. I forgot to subtract the window size from startrows. It should be:

startrows = 1:30:size(y, 1)-windowsize+1;

same as in the ndgrid code.

As for your later code. You have loaded the data into a variable called data but then use date which happens to be a matlab function. Note that as I've replied, if you want to apply genhurst by month, the easiest is with:

data = readtable('data.xlsx');
monthly_data = groupsummary(data, 'Date', 'month', @genhurst)

or with retime.

[BISWA BHUYAN]

I am sorry that i assume that i could not make you understand, may be due to lack of my clarification for my second query. Let me to explain again,

I applied the suggested code to a 3 column matrix without date column and it works very well

windowsize = 100;

startrows = 1:30:size(y, 1)-windowsize+1;

HH = cell(numel(startrows), size(y, 2));

for col = 1:size(y, 2)

for row = 1:numel(startrows)

HH{row, col} = genhurst(y(startrows(row):startrows(row)+windowsize-1, col));

end

end

Now, my second query was, i want to apply the above code where the equal number of date column will be in matrix. Then, I want to store the results with date in a new matrix, where the first window result will be store in the "windowsize+1" row with the corresponding date in a new matrix. For example, in the atatched data, the result of the first window of the x1 variable (assumed 0.25) and the second variable x2 (assumed 0.56) will be stored as follows,

[ 23-05-1994 0.25 24-08-1990 0.56]

Because, the "windowsize+1" row date of x1 is 23-05-1994 and the "windowsize+1" row date of x2 variable is 24-08-1990. Similarl procedure will be followed for all rolling windows. I have atatched the data file.

Please suggest

Coming to the suggestion of the monthly genhurst, i am reading your suggestion on groupsummary and retime. Then I will use code, if any problem i will let you know in future.