On Jul 10, 6:31 pm, Sudeep <sudeepkumarj...@gmail.com> wrote:
On Jul 10, 1:55 am, Ray Koopman <koop...@sfu.ca> wrote:
On Jul 9, 9:16 am, Bruce Weaver <bwea...@lakeheadu.ca> wrote:
On Jul 9, 9:01 am, Sudeep <Sudeepkumarj...@gmail.com> wrote:
**I will try my best to keep the discussion generic from a statistical
point of view and not get too involved with creating a communication
barrier wrt to my research area.**
Collected Sample of n=139 (M&A deals between years 2001 - 2006)
Calculated CAR(Cumulative Abnormal Return) for all 139 deals post-date
of deal announcement.
Sample divided into three groups (viz. Cash, Stock, Cash+Stock based
on the method of payment used in each deal)
Cash count =71 = n1
Stock count = 43 = n2
Cash+Stock count = 25 = n3
Total = 139 = n
Trying to disprove null hypothesis (H0): Cash deals on average have
not yielded positive abnormal returns.
***Struggled with using parametric statistical methods to disprove H0
***Conducted an initial study into non-parametric statistics and found
Kruskal-Wallis test (H Statistic), Friedman, and Bonferroni-Dunn
method to be suitable.
Need help with selecting the right method with possible correction for
significance error.
Any help will be greatly appreciated!
Sudeep, when posting the same thing to more than one group, please
post one message with all the groups listed on the Newsgroups: line.
This ensures that all responses are kept together, which makes things
much nicer for anyone searching the archives in future. It also
allows folks with certain newsreaders to view your message only once
if they happen to read both groups.
Note to other group members: If you want to respond to Sudeep's
question, please respond to this message, which is cross-posted to
both groups.
--
Bruce Weaver
bwea...@lakeheadu.cahttp://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."
The sample sizes are large enough that the sampling distributions of
the sample means should be close enough to normal to let you use the
Brown-Forsythe one-way anova and and ordinary heteroscedastic t-tests.
Thank you Ray
I am trying to compare the three groups against each other and hail
the one with the most positive abnormal returns to be the best method
of payment to buy a company. The null hypothesis I posted earlier is
only 1 of 3.
I did some reading on Heteroskedasticity... if I can use either the
Levene test or the Brown-Forsythe test to confirm the sample to be
heteroscedastic then I can go on run a heteroscedastic t-test? Is that
what you meant to convey? or am I going astray here?
Sudeep
Apologies
I failed to mention earlier... the 3 classes/groups are non-
parametric.
Sudeep
