Date:         Mon, 14 May 2012 15:29:29 +0200
Reply-To:     =?windows-1252?Q?Marta_Garc=EDa-Granero?=
              <mgarciagranero@gmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         =?windows-1252?Q?Marta_Garc=EDa-Granero?=
              <mgarciagranero@gmail.com>
Subject:      Re: Tennis
In-Reply-To:  <1336997214883-5708910.post@n5.nabble.com>
Content-Type: text/plain; charset=windows-1252; format=flowed

Hi:

I think you have paired data, since in the first match, one member of
the pair wins, and, therefore, the other one loses. You have to take
into account the pairing and treat the data as a matched cohort study.
McNemar test will do that, but you get only a p-value, not an estimate
of the increase of likelihood of victory. You need a relative risk that
takes into account the matching. Look for this paper: The Stata Journal
(2004) 4, Number 3, pp. 274–281, Analysis of matched cohort data, by
Peter Cummings Barbara McKnight.

I have also read that log-binomial or Poisson regression can be used for
that.

I hope this helps,
Marta GG

El 14/05/2012 14:06, drfg2008 escribió:
> As far as I understand a Chi-square test is testing independence between the
> two sets, which is definitely not the case since the winner of set 1 is
> stronger and will be more likely to win set 2, too. And this is not my
> question. The question is if the likelyhood increases to win again when the
> player has won before.
>
> I thougth about the McNemar test first since this tests if a change from
> winning to loosing is equal likely as the change from loosing to winning. If
> the hypothesis (see above) is true (namely that a winner is more likely to
> win again) there should be more changes from one side to the other.
>
>
>
> -----
> Dr. Frank Gaeth
> FU-Berlin
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Tennis-tp5708820p5708910.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD