One Sided T test function gives -Inf as part of the confidence interval?

As the question states, I have a problem where I need to use the 'ttest' function.

I have the data,

data = [37 , 54 , 29 , 46 , 33];

then I run the test:

[h , p , ci , stat] = ttest(data , 45 , 'Alpha' , 0.10 , 'Tail' , 'left');

I get:

h = 0; p = 0.158; ci = -Inf 49.4618; stats = tstat: -1.1474; df: 4; sd: 10.1341;

Everything is correct from hand calculations, except for the -Inf value as the lower limit on the confidence interval. I looked for help with this before posting, and a previous post (<http://stackoverflow.com/questions/3816702/ttest-function-error-in-matlab-error-using-nanmean-too-many-input>) about an unrelated issue with ttest suggested using "which" to verify the correct function is being used.

When I do that, I get:

C:\Program Files\MATLAB\R2013a\toolbox\stats\stats\ttest.m

I don't know if this is a flaw with the function or something else. The script is attached.

Answer 1

[the cyclist]

The result is correct. You have asked for a one-sided test. The null hypothesis is that the mean of the data is 45. The alternate hypothesis is that the data come from a distribution with mean less than 45.

The test did not reject the null (h =0), and the 90% confidence interval is that the mean is less than 49.4618, which equivalent to saying it is in the interval [-Inf,49.4618].