How to select rows that precede and follow a specific row with a certain value?

Hey there,

I'm currently working with experimental data that is stored in a 288 x 25 double (288 = total number of trials ; 25 = variables of interest). Column '13' represents the 'correctness' coded as 1 = correct and 0 = error. What I'm trying to do is to select sequences of '1 0 1' that is every row that precedes an error trial and every row that follows the same error trial. I want to store the 'pre-error-trials' and the 'post-error-trials' as variables separately.

Since I'm still a newbie to Matlab I'm not familiar with every concept but I guess that logcial indexing might play a role here?

I would be very grateful if you could provide some help.

Answer 1

[Matt J]

c=data(:,13);

 preError=data( c(1:end-1)==1 & c(2:end)==0 ,:);
postError=data( c(1:end-1)==0 & c(2:end)==1 ,:);

[Steve Schulz]

Hey Matt,

thanks for your help!

While the 'preError'-command is working fine I don't see why the 'post-Error'-code gives me the error itsself instead of the correct trial that follows an error.

[Matt J]

Try this,

postError=data( find(c(1:end-1)==0 & c(2:end)==1)+1 ,:);

[Steve Schulz]

This one does the trick!

But how can I extract sequences such as '1 0 1' ? The overall aim of this approach is to select every pre-error-trial with it's corresponding post-error-trial.

Linking 'preError' and 'postError' gives me an error message: 'Error using ''&''

Arrays have incompatible sizes for this operation.'

Thanks a lot so far, I really appreciate your help!

[Matt J]

That's a little different. Note that it is not necessary for a pre-error trial to have a corresponding post-error trial, nor vice versa. If you only want to restrict attention to those that do, then,

c=data(:,13);
idx=  find( c(1:end-2)==1 & c(2:end-1)==0 & c(3:end) )+1; ;

preError=data(idx-1 ,:);
postError=data(idx+1,:);

[Steve Schulz]

Oh, it seems that you are familiar with this? Yes, you're right, it's not obligatory that a pre-error trial has a corresponding post-error trial but this approach can yield to less biased results therefore I wanto to give it a shot.

Thanks, this helps me a lot!

[Matt J]

Oh, it seems that you are familiar with this?

No, but clearly if the column data(:,13) begins or ends with a 0, it will happen.

[Steve Schulz]

Ah, I got it. Thanks !