Identifying if reaction times are 3 standard deviations away from mean

have a data set of several thousand values.

There are 10 unique subj_indx, 1 through 10. Each one represents an experimental subject. Each subject has a bunch of reaction times, rt.

I am new to Matlab and don't understand how to write for loops. I know I need to run through the entire dataset, and identify outlier reaction time trials, without using isoutlier function. This is because I want to be ale to identify whether each RT trial is > 3 standard deviations away from the mean RT, and then mark that trial in a new vector as "1", otherwise, mark it as "0."

I know Matlab has standard deviation and mean functinos, but am not sure how to connect everything together for this code.

I don't know where to start other than:

rt=subject_data.rt;

for id = 1:length(subj_idx)%loop for each unique id value

%rtIdx = 1:length(rt)

currentid = rt(rtIdx);

rtIdx = 0;

and I'm not sure even that's correct.

[Steven Lord]

My first thought would be to use the isoutlier function but you asked how to do it without that function. Is there a reason you can't use it?

[Rania Hanna]

I want to check if each trial is 3 std dev from the mean, so I don't think isoutlier will work with that.

[Steven Lord]

See the description of the "mean" method on that isoutlier documentation page to which I linked.

[Rania Hanna]

Perfect, thanks so much!

Answer 1

[Chunru]

% rt=subject_data.rt;
rt = randn(300, 1) * 3 + 1;         % genate some random data
mu = mean(rt)

mu = 1.2843

sigma = std(rt)

sigma = 3.0914

idx = abs((rt-mu)) > 1*sigma;       % use 3 sigma when samples are large
[rt idx]

ans = 300×2

2.6351 0 -0.5263 0 -1.5895 0 2.5097 0 4.4268 1.0000 1.5017 0 -3.1915 1.0000 -1.7580 0 1.9555 0 1.5195 0

[Rania Hanna]

Thank you!

Answer 2

[Steven Lord]

Take a look at the "mean" method for the isoutlier function.

[Rania Hanna]

Thank you!