 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: RE: ttest, signrank, ranksum with for each?
The tests you mention are all based on an assumption of independent observations. Even though you are looking at differences, the question still remains whether that is consistent with your ideas about data generating process(es)?
Nick
[email protected]
René Geppert
I have data for 2 groups. For the first group I have 6 quarters of
observations, for the second group I have 2 quarters. 1. Q 2007 - 2. Q
2008, and 1. Q 2008 - 2. Q 2008 respecively.
Data are in long format, like this.
Company - Year - Quarter - Variable1 - Group
abc corp - 2007 - 1 - 0.4 - 1
abc corp - 2007 - 2 - 0.1 - 1
abc corp - 2007 - 3 - 0.05 - 1
abc corp - 2007 - 4 - 0.11 - 1
abc corp - 2008 - 1 - 0.70 - 1
abc corp - 2008 - 2 - 0.0 - 1
def corp - 2008 - 1 - 0.1 - 0
def corp - 2008 - 2 - 0.8 - 0
ghi corp...and so on
I now want to perform difference in means and difference in medians tests
for 'Variable1' for both, changes from quarter to quarter within each
group (0 and 1) and differences among the groups within one quarter.
Similarly, I want to test for difference in medians using signrank for
differences within one group across quarters and within one quarter
across groups using ranksum.
So: What I basically want to examine is, have there been any significant
changes in variable 1 between the quarters and between the groups?
Because multiple observations exist for 'Quarter', I cannot simply do a
command like:
. ttest Variable1, by(Group) cuz this captures all the quarters at once
I tried nested if-criteria which doesn't work.
So I presume this might be possible using a for each command, anyone got
an idea?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/