Most testing should be considered 2-tailed. The exception is if you can
honestly say yes to either of the following questions:

Before looking at the data, did you know which group MUST be larger if there
was a difference?

Before looking at the data, were you only interested in a difference if it
was in a particular direction (and it then turned out to be in that
direction)?

p=0.087 means that there is less than a 1 in 10 chance of observing such an
extreme difference (in the pre-specified direction) simply by chance if in
fact there is no difference. It is not significant at the 0.05 level, but
depending on your objectives, may be worth pursuing.

Jerry

"gbh" wrote:

Can you help me figure out if I'm interpreting the t-probability correctly?
I am comparing the means of 2 samples. With a 2-tailed test the probablility
statistic is 0.173 & with a 1-tailed it is 0.087. My interpretation is that
there is not a signigicant difference between the means using either the 1 or
2 tailed test. To be honest, I don't have a clue whether I should have used
a 1 or 2 tailed test so I did both. Can you also tell which I should be
using?
In another comparison of the means of 2 samples, the probability statistic
was 0.000346. I am interpreting this as a significant difference at the .03
CI. Is this correct?
gbh

"Alex" wrote:

Anita

The TTEST function is a statistical test known as the t-test.

Put simply, it compares the means of two samples to assess whether they
differ significantly or not. The output is a probability value. In general a
value less than 0.05 would indicate that the two means are different.

For example, suppose you had two classes of schoolchildren, each of class
size 30. Suppose they all took the same test and received a mark out of 100.
A question may be to test whether the average (mean) mark for one class is
significantly different than the other class.

Class A ClassB
87 82
92 77
63 75
70 73
65 87
etc

In Excel you may have column A1 as class A results and column B1 as class B
results.

The mean of class A maybe 80 and the mean of class B maybe 75 ( I am making
these numbers up). So, does this suggest that the average results are
signifcantly different?

What you would do is...

TTEST([input range for classA],{input range for classB], 2,2)

And this will give a probability value e.g. 0.15. This means that the chance
of observing those results would occur about 15% by chance anyway. Typically,
a p-value of less than 0.05 (5%) is considered significant i.e. those results
would only be obtained by chance 5% of the time.

This is a very brief overview into a subject that has many nuances. This is
a light overview. If that satisfies your curiousity then fine...if not then
please do write back and I shall endeavour to explain further.

Regards

Alex



"Anita" wrote:

Hello

Can someone explain what this function does? I have to do a support call
for a customer and have never come across the Ttest function before. The
answer that I have in the cell containing the function has got E-15 after it
(which I don't understand either). I've also been asked how to calculate
probability. The version is 2000 - I can supply more detail if the above
isn't adequate.

Many thanks for your help.

Anita