Confidence interval (Single sample t-test)


I am afraid I misunderstand the results of the Single sample t-test. For example:

Iris dataset, second numeric column (sepal width) with 3.0 as test mean (measured: ~3.054). I get:

% CI (Lower) CI (Higher)
90 -0.0046 0.1126
95 -0.0160 0.1240
99 -0.0384 0.1464






Should not those get narrower and narrower intervals? Isn't it usual to report the absolute values not the relative to the test value?

Thanks, gabor


PS: Sorry, for some reason the table is not visible as it is in the editor window. I have made some emphasis in the data (header and the CI in % values)

I would expect the results to get broader and broader, because to be 99% confident, you would require a broader range, whereas you can be less confident at 90% say, that the results fall within a narrower range.

I would assume the data shown is the relative results, in that this range is how far above and below the mean value of 3. i.e. that at 99% confidence, the true mean lies between 2.9616 and 3.1464. That is how I understand the data.


Thanks Simon.

I guess I should think more about it when I was not tired. :) Now I do not understand how I made that conclusion. I should just have created a figure.

Cheers, gabor