Best way to seperate a very large array of numbers into different variables based on a value tat is given in the first column?

I have a variable that contains experimental data, it is 96x300001. (however the size will change every time I run my script). The first column gives a value which represents a certain group (e.g group 1, group 2 group 3 etc. The variable kinda looks like this =

1 0.021 0.0433 0.656, 
2 0.00321 0.335 0.605, 
3 0.1 0.035 0.625. etc etc

I want to tell MATLAB "numbers next to the number 1 go into variable "group1" and numbers next to the number 2 go into a variable called "group2" and so on, but I have no idea how to do this!

[Marissa Holden]

For anyone looking to answer this question in the future I ended up using the "find" function. "data" is the variable containing my data -

idx_group3 = find(data(:,1)==3); -- this will output a variable with the number of each row that starts with the number 3.

Then, I tell matlab to put the data from those particular rows (from 2 to the end of the array) into a different variable.

data_group3 = data(idx_group3,2:end);

Answer 1

[Cedric]

Here is an example: we build a fake data set:

 >> data = [sort(randi(3,10,1)), rand(10,5)]
 data =
    1.0000    0.9727    0.6504    0.8675    0.9329    0.2748
    1.0000    0.9217    0.1238    0.1572    0.7492    0.7884
    1.0000    0.7864    0.9880    0.6371    0.8526    0.8418
    1.0000    0.2187    0.6194    0.9931    0.9952    0.9586
    2.0000    0.1218    0.9761    0.8390    0.6048    0.2233
    2.0000    0.5117    0.0724    0.9759    0.2118    0.7046
    3.0000    0.6800    0.3582    0.4349    0.8793    0.8167
    3.0000    0.8622    0.4866    0.7729    0.8863    0.8443
    3.0000    0.6063    0.8812    0.5100    0.6926    0.9223
    3.0000    0.7486    0.0681    0.4883    0.6132    0.4820

and we group it this way:

 >> groups = splitapply( @(x){x(:,2:end)}, data, data(:,1) )
 groups =
  3×1 cell array
    {4×5 double}
    {2×5 double}
    {4×5 double}

Now groups is a cell array that we can index using the group ID:

 >> groups{1}
 ans =
    0.9727    0.6504    0.8675    0.9329    0.2748
    0.9217    0.1238    0.1572    0.7492    0.7884
    0.7864    0.9880    0.6371    0.8526    0.8418
    0.2187    0.6194    0.9931    0.9952    0.9586

 >> groups{2}
 ans =
    0.1218    0.9761    0.8390    0.6048    0.2233
    0.5117    0.0724    0.9759    0.2118    0.7046

 >> groups{3}
 ans =
    0.6800    0.3582    0.4349    0.8793    0.8167
    0.8622    0.4866    0.7729    0.8863    0.8443
    0.6063    0.8812    0.5100    0.6926    0.9223
    0.7486    0.0681    0.4883    0.6132    0.4820

[Marissa Holden]

it's saying that splitapply is not a function? - I am using 2014b

[Image Analyst]

Introduced in R2015b.

[Marissa Holden]

Is there a similar function in 2014b? Or a way I can use this in my version?

[Cedric]

Yes, try the following instead:

 groups = mat2cell( data(:,2:end), accumarray( data(:,1), 1 )) ;

If you evaluate

 accumarray( data(:,1), 1 )

you will see that it computes the length of each group, and then reading the doc of MAT2CELL, you'll understand what we do here.

Answer 2

[Image Analyst]

You may not need to do that. Whether you do or not depends on what you're going to do next. For example if you just wanted the means or other simple stats, you could use grpstats() without creating variables group1, group2, group3, .... groupn. Or if you just wanted to loop over all groups, you could just extract rows one group at a time into a single variable "thisGroup" and then process that within the loop. So, assuming you know what groups each row belongs to (and you do already know this), what do you plan on doing next?

[Marissa Holden]

no I 100% need them separated unfortunately

[Image Analyst]

Can still you answer my question about what you want to do assuming you had them separate?

[Marissa Holden]

The values correspond to a raw data trace from a cell. I want to plot the mean value of the data trace over time.

So I would want to plot the mean from all the values that are next to the number 1 and so on.

[Cedric]

Your reference to GRPSTATS may be useful to me ImageAnalyst (I'll test it against my current method, in a function for performing zonal statistics), thanks!

[Image Analyst]

Marissa: Do you have the Statistics and Machine Learning Toolbox? This is exactly what it does, with no need to create a bunch of separately named variables!

[Marissa Holden]

I'll have to install it! Thanks for letting me know, it looks very useful to me :)