Indexing and filling matrices?

Hi again,

I have the following code to compute the Kaplan-Meier estimates with right-censoring:

T = size(time);

n = zeros(T(1),T(2));
d = zeros(T(1),T(2));

KMadjRank = zeros(T(1),T(2));
D = zeros(T(1),T(2));
N = zeros(T(1),T(2));

for i = 1:1000 %right for 1 column

      if censoring(i) == 1 % if failed, increase number of failed by 1

          d(1) = 0;
          d(i+1) = d(i)+1;

      end

      if censoring(i) == 0 % if censored, use previous # of failed

          d(i+1) = d(i);

      end

      n(1) = T(1); % Initialize number at risk as sample size

      n(i+1) = n(i) - 1; % decrease number at risk by one

end

for i = 2:999

      KMadjRank(1) = 0; % Initialize Prob of survival between 0 and first failure

      KMadjRank(i) = (1-(d(i-1)/n(i-1))).*(1-(d(i)/n(i))); % KM product-limit estimator

end

So I have a 100x1000 matrix of times and of censoring and I'm trying to get KM right. Right now, this code fills out the first column correctly but I don't know how to get it to correctly index through column by column instead of as the entire thing. I want to treat this as 1000 samples of 100 so I'm trying to compute the Kaplan-Meier estimates by column so for the individual samples but I'm struggling with the coding language.

Can anyone help?

Thanks!

[Erin Winkler]

I now have the following which reads the error "Array indices must be positive integers or logical values" and I'm not sure where that's happening. But I have this now (which is more correct) but still has the same issue as before:

function [KMadjRank] = AdjustedKMrank(time,censoring)
%sampleSize = T(1)

timesort = sort(time);

T = size(time);

n = zeros(T(1),T(2));
d = zeros(T(1),T(2));

KMadjRank = zeros(T(1),1);

for i = 2:99
    t = timesort(:,i);

      if censoring(t) == 1 % if failed, increase number of failed by 1

          d(1) = 0;
          d(i+1) = d(i)+1;

      end

      if censoring(t) == 0 % if censored, use previous # of failed

          d(i+1) = d(i);

      end

      n(1) = T(1); % Initialize number at risk as sample size

      KMadjRank(1) = 1; % Initialize Prob of survival between 0 and first failure

      KMadjRank(i+1) = (1 - d(i)/n(i))*(1 - d(i+1)/n(i+1)); % KM product-limit estimator

      n(i+1) = n(i) - 1; % decrease number at risk by one

end

end

[David Fletcher]

You don't say where the error is occurring. What sort of data is contained in timesort? If it is anything other than positive integers or logicals, then you will get a problem with using it as an index (for example here)

    if censoring(t) == 1

[Erin Winkler]

It does not indicate where the error is occurring, which is why I asked. It does not give me a specific line number.

It looks like time and censoring are doubles but I tried changing them using:

time = logical(time);

but that didn't help.

[David Fletcher]

I've not been using Matlab all that long, but I have never known it not give a line number from where the error is occurring (at least not when running a script file/function). Quite often you'll get multiple line numbers from where it does a stack trace. To be honest, in that code you've posted that's the only obvious place that I can see that will give an error like that, but without a line number I don't even know if the error is coming from that function code or whatever is calling that function

[Stephen23]

@Erin Winkler: please give us the complete error message. This means all of the red text. Really, all of it. Do not leave any out.