How to make matrix dimensions the same size?

I have data which currently is a fixation report. This means it has starting fixations (SF) and end fixations (EF). For example, in one row the fixation has an SF of 7 and an EF of 50. So one row is 1 fixation with the start and end time.

I want to turn this data into one row per milisecond (in my case there is 39500msec so 39500 rows). With a 1 if there is currently a fixation and 0 if there isnt. With this per participant (PPT).

I have tried to following but get the error message that 'Matrix dimensions must agree'.

I am new to Matlab!

data = readtable('data.txt','HeaderLines',1,'Delimiter','\t');
    PPT=data.Var1;
    clip=data.Var2;
    condition=data.Var3;
    SF=data.Var4;
    EF=data.Var4;
    OnSpeaker=data.Var6;
    
UP = unique(PPT);
jj=1:39500;
timeMsec=transpose(jj);
 
for x=1:length(UP) %loop through all the unique participants
for z=1:length(OnSpeaker) %loop through the length of the data

timeseries=zeros(39500,1);
end   
    
for f = 1:length(SF); 
if SF>= timeMsec & EF<=timeMsec    % This is definitely the bit thats wrong!!
timeseries=1;
else timeseries=0,
end
end

[Walter Roberson]

Please give an example of some lines in the text file.

[Jessica Dawson]

Thank you!

I want one line per msec rather than per fixation!

[Jessica Dawson]

I am also adding an example of what I would like the output to look like!

Answer 1

[neil jerome]

% fixation time script

% read data
data = readtable('data.txt','HeaderLines',1,'Delimiter','\t'); % see my 'data.txt' below
PPT = data.Var1;
clip = data.Var2;
condition = data.Var3;
SF = data.Var4;
EF = data.Var5;
OnSpeaker = data.Var6;
UP = unique(PPT); nUP = length(UP);

maxTime = 20; % end recording time in ms
% assign matrix: rows are time, cols are UP
dataMat = zeros(maxTime, nUP);

%%
for aa = 1:nUP % loop for unique participants
    thisUP = UP{aa};
    thisUPdata = strcmp(thisUP, PPT); % find which rows for this UP
    for bb = 1:length(PPT) % work through line by line
        if thisUPdata(bb) % if this line if for the current UP
            dataMat(SF(bb):EF(bb)-1, aa) = 1; % assign ones for fixation
        end
    end
end

%% add time column if you want to
fullTable = horzcat((1:maxTime)',dataMat);

%% here is the data.txt, modified to show multiple UP, and much shorter :)
% PPT	Clip	Cond	SF	EF	OnSpeaker?
% ppt15a	Group1_Clip1_NaturalX.xvd	Natural	2	5	0
% ppt15a	Group1_Clip1_NaturalX.xvd	Natural	7	9	0
% ppt15a	Group1_Clip1_NaturalX.xvd	Natural	12	14	0
% ppt18a	Group1_Clip1_NaturalX.xvd	Natural	3	9	1
% ppt18a	Group1_Clip1_NaturalX.xvd	Natural	10	12	1
% ppt18a	Group1_Clip1_NaturalX.xvd	Natural	15	16	1
% ppt20a	Group1_Clip1_NaturalX.xvd	Natural	4	19	1

[Jessica Dawson]

You are a genius! Thank you so much! And thank you for explaining/commenting everything so clearly for me to understand, it is really appreciated.

[neil jerome]

you're welcome; good luck :)

[Jessica Dawson]

Could I also ask, if I want to know in addition to this. When dataMat is 1 AND the onSpeaker variable is a 1,0 or -1. I have tried the below but it isnt correct.

Thank you again!

% read data
data = readtable('data.txt','HeaderLines',1,'Delimiter','\t'); % see my 'data.txt' below
PPT = data.Var1;
clip = data.Var2;
condition = data.Var3;
SF = data.Var4;
EF = data.Var5;
OnSpeaker = data.Var6;
UP = unique(PPT); nUP = length(UP);
maxTime = 39500; % end recording time in ms
% assign matrix: rows are time, cols are UP
dataMat = zeros(maxTime, nUP);
dataSpeak=zeros(maxTime,nUP)
 
for aa = 1:nUP % loop for unique participants
    thisUP = UP{aa};
    thisUPdata = strcmp(thisUP, PPT); % find which rows for this UP
    for bb = 1:length(PPT) % work through line by line
        if thisUPdata(bb) % if this line is for the current UP
            dataMat(SF(bb):EF(bb)-1, aa) = 1; % assign ones for fixation
        end
    end
    % So this gives me a table of whether there is or isnt a fixation per
    % ppt
    
    % Now i want to know whether there IS a fixation (dataMat = 1) but also, if
    % the 'OnSpeaker' variable is a 1,0 or -1)
    
    for bb=1:length(PPT);
        if thisUPdata(bb);
            if OnSpeaker(bb)==1 & dataMat(bb)==1
                dataSpeak(bb)=1;
            end
            if OnSpeaker(bb)==0 & dataMat(bb)==1
                dataSpeak(bb)=0;
            end
            if OnSpeaker(bb)==-1 & dataMat(bb)==1
                dataSpeak(bb)=-1;
            end
        end
    end
    
end
 
 
% add time column if you want to
fullTable = horzcat((1:maxTime)',dataMat);
speakTable = horzcat((1:maxTime)',dataSpeak);

[neil jerome]

different ways to do this, so what is 'best is very subjective/dependent on what you're doing next. and that's assuming i understand what you mean :)

what you have here looks ok to start, but:

you only have a one-dimensional dataSpeak, which either implies that the speakerOn condition is the same regardless of participant (ie all participants get the same timings), or you're overwriting this every time you loop to the next UP. so if this would be bad, you need to make it a matrix with size that matches dataMat where you can assign a separate dataSpeak column for each UP. so you have corresponding entries in dataMat and dataSpeak (or you could combine these into a single 3D matrix, with the first layer fixation and the second layer speakerOn; this is more compact, but less 'readable'):

for bb = 1:length(PPT) % work through line by line
        if thisUPdata(bb) % if this line is for the current UP
            dataMat(SF(bb):EF(bb)-1, aa) = 1; % assign ones for fixation
            
            % extra line that assigns speakerOn in the same indices as you have the fixation
            dataSpeak(SF(bb):EF(bb)-1, aa) = OnSpeaker(bb); % assign [-1,0,1] for speaker
        end
end

although this would give you 'valid' dataSpeak values only where there were fixations, and so you could initialise the array as NaNs, not zeros, so you don't get any 'default zeros' that weren't explicitly put there by reading the data file. (though beware using NaNs in tables when you do any other maths on the table, you need to use nansum() instead of sum(), etc)

hope that makes sense; if i've misunderstood, sorry. but i have assumed that you only really care about the speaker when you record a fixation, otherwise you would have some separate listing of speakerOn that wasn't linked to fixation events?

couple of notes on what you wrote: since OnSpeaker has the value you will write to your data matrix, you don't need to go through testing for each value and then explicitly assign those same values! so you can save space/time:

            if OnSpeaker(bb)==1 & dataMat(bb)==1 % use double: && as the boolean 'and' operator
            dataSpeak(bb)=1;
            end
            if OnSpeaker(bb)==0 & dataMat(bb)==1
                dataSpeak(bb)=0;
            end
            if OnSpeaker(bb)==-1 & dataMat(bb)==1
                dataSpeak(bb)=-1;
            end
            
            % all this becomes
            
            if dataMat(bb) % no need to write '==1' the 'if' will return true for value of 1
                dataSpeak(bb) = OnSpeaker(bb);
            end
            

ok, back to work.

[Jessica Dawson]

This all makes sense completely! Thank you so much for your time and explaining so clearly! I have learnt more from you than a weeks Matlab course I took!!

It is really appreciated!