t test in Excel

[coffee@work]

Hello,

In a data set I have to distinguish trends. We think that early
adopters of a product consume more than late adopters. Therefore I
would like to use a t-test to see if their average consumption is
different(if anyone has another idea, you are welcome). I'm used to
work with SPSS but here only Excel is available so that I experience a
few problems. In a SPSS you have two sample t-test and the results are
there. With excel you have to choose a lot things.

I have chosen tools >>> Data Analysis and then two sample assuming
equal variances (when equal and when unequal?). The data has to be
inputted but than you have to fill in the hypothesized Mean
Difference. I filled in 0 because I think the difference between the
means has to be 0 if I want to test if means are different. I am
right? Which other option can be useful using this tool?

Kinds regards,
Bjorn

 

[Alex]

Bjorn

In your example, using 0 as the hypothesised mean is
correct becuase your null hypothesis is that there is no
difference in product consumption between earlier
adopters and late adopters.

As for variance, if you have good reason to think that
the variance between the the two data sets will vary then
you should not assume equal variances. When you do the
data analysis and get and Excel output, it will show the
variance for the the two samples. If they are similar
then you know you were right to assume equal variances.

Also in the output, you probably want to use the Two-
Tailed T-Stat. In your example, unless you are looking
for a bias in one direction for consumption, then it is
best to use a two-tailed test. In general this is good
practice unless you have an excellent reason to believe
that an effect wil only ever be observed in one direction.

If it helps, Snedecor 'Statistical Methods' is a great
book on statistics as a readable introduction. It covers
two-sample t-tests etc.

Hope this helps. If you have any other questions please
do ask. I shall be at my computer until 14.00GMT

Thanks


Alex