How do I do SEM Error Bars?

I know how to find the standard deviation with error bars but how do I find the standard error mean of variables?

Answer 1

[Star Strider]

You have to know the number of observations that went into calculating each standard deviation.

Assuming ‘x’ is a vector:

SEM = std(x)/sqrt(length(x));

[Delany MacDonald]

So if it's a 21 x 345 array and I find the mean over the 21 days would it be A =21 x 345 SEM = std(A)/sqrt(length(A));

[Star Strider]

Not quite. You have to change that a bit to calculate along the rows:

A = randi(99, 21, 345);                                                 % Create Data
SEM_A = std(A, [], 2)./ sqrt(size(A,2));                                % Calculate Standard Error Of The Mean
CI95 = bsxfun(@plus, mean(A,2), bsxfun(@times, [-1  1]*1.96, SEM_A));   % 95% Confidence Intervals

I tossed in the 95% confidence interval calculation for free!

[Vlatko Milic]

Hi,

I get two outputs when calculating the standard error according to your example. What can the reason for this be?

Kind regards

[Star Strider]

My code calculates ±95% confidence limits, so it produces both limits.