How to calculate the mean for the number of occurances

suppose i have a cell C where

C{1,1}= ( 2 2 2 2 3 3 4 4 5);
C{2,1}=(3 4 4 4 5);
C{3,1}=(4 5);

and so on

i want the output as

x = (2 2 2 2 3 3 4 4 5);

the output should be the value which occurs the mean number of times of the occurances in the input. if the value is a decimal then it can rounded to a greater whole number for ex: the value 3 occurs twice in first cell and once in second cell. 2+1/2= 1.5 can be rounded to 2. so it occurs twice in the output.

Answer 1

[dpb]

u=unique([C{:}]);
n=cell2mat(cellfun(@(x) histc(x,u),C,'uni',0));
n(n==0)=nan;
x=cell2mat(arrayfun(@(x,n) repmat(x,1,n),u,round(nanmean(n)),'uni',0));

>> x =
     2     2     2     2     3     3     4     4     5
>> 

[johnson saldanha]

Error using horzcat

Dimensions of arrays being concatenated are not consistent.

Error in accelerationrate (line 25)

u=unique([C{:}]);

im getting this error.

[dpb]

Need to see what you are using specifically for C, then...works fine for a 1D cell array of double vector arrays for me--

>> C
C =
  3×1 cell array
    {1×9 double}
    {1×5 double}
    {1×2 double}
>> ([C{:}])
ans =
     2     2     2     2     3     3     4     4     5     3     4     4     4     5     4     5
>> unique([C{:}])
ans =
     2     3     4     5
>> 

The colon operator inside the [] to catenate the list returns all elements of C as a row vector and unique subsequently returns the subset vector.

How is your C cell array different than above?

NB: If don't make the orientation of the vectors you incorporate into C consistent, however, that will cause such an error. If take the above and add

>> C{4,1}=[4 5].';
>> u=unique([C{:}]);
Error using horzcat
Dimensions of matrices being concatenated are not consistent. 
>> 

The problem here is, of course, that the last cell is a column vector instead of a row vector.

>> C
C =
  4×1 cell array
    {1×9 double}
    {1×5 double}
    {1×2 double}
    {2×1 double}
>> 

so the input doesn't match the expectations. In that case you would need to write

u=unique(cat(1,C{:}));

to catenate vertically; the [] Matlab idiom does catenation of the list elements horizontally and when there are differing number of elements on a row, that fails.

[johnson saldanha]

all are column vectors. and i want to run them inside a loop. coz its not just 3 cell elements. there are more so for i=1:length(C)

[dpb]

It doesn't make any difference how many there are; you simply have to ensure you use the correct orientation for the cell vectors for whichever syntax you use to catenate them.(*)

If they're row vectors then [] works, it column then cat(1,.. Of course, you can replace [] with cat(2,... for the row vector case as well.

(*) Or, probably more correctly stated, use the correct concatenation syntax for the orientation of the cell vectors.

NB: However, recognize that if you work with column vectors, then the orientation of the unique vector is also going to be column wise and then the n array will be transposed so the nanmean function will need to operate along the row instead of default dimension...

Making the change up front has ramifications all the way through.

[johnson saldanha]

Error using arrayfun

All of the input arguments must be of the same size and shape.

Previous inputs had size 16 in dimension 1. Input #3 has size 1

Error in accelerationrate (line 29)

x=cell2mat(arrayfun(@(x,n) repmat(x,1,n),u,round(nanmean(n)),'uni',0));

im getting this error now

[johnson saldanha]

any idea why im getting this error?

[dpb]

Yeah, again (still?) you've got one array as a row vector (#3) while the previous two were column vectors (size 16 in dimension 1).

You must pick an orientation for your cell array content organization and be consistent--otherwise, you'll have to write all kinds of patchwork code to correct such orienation mismatches going forward.

[johnson saldanha]

all are column vectors. theres no row vector

[dpb]

That's not what the error message says... :)

If the vectors inside the initial cell array are columns instead of rows as your initial example, then as above you'll have to first cat(1,...) them, but then the counts array will end up being a column vector instead of the 2D array because cellfun will just concatenate those columns into one long (column) vector instead of concatenating the rows vertically and you've thus lost the orientation that lets nanmean know how to average across the cells as is the whole end objective.

Probably the easiest way to code it would be to transpose the argument of the histc function inside the anonymous function so it is again a row vector there as needs must be...otherwise, you've got much more painful gyrations to build the necessary array shape.

Step through each step at the command line for a (smallish) example so you can see the results of each operation and get the proper orientation for the job.