Dear All,
I have a binary variable (presence or absnece of an event) of around
500 cases. I have been asked to provide the value of the mean along
with an upper / lower bound of this dichotomous variable. For this
reason, I computed the mean and the confidence intervals of the
variable in question.
In addition, i did bootstrapping so as to raise the precision of the
arising results. Having successfully done it, I asked STATA to
bootstrap the statistics of the mean and its confidence intervals
along with the standard error of the mean. Stata responded by
producing the observed point estimate for each of these statistics
along with the confidence intervals for each of the computed bootsrap
statistics.
As i have very limited knowledge of the bootstrapping methodology,
I would be grateful to you if I would receive any suggestions and
comments about the kind of numbers which I would need to report. For
instance, could it be suffice to report the observed mean along with
its upper/lower bound at the 95% singnficance level as it has been
computed by STATA? Would it be equally acceptable to ignore the
bootstrap methods and present the means and the CI figurs as produced
by the CI command? Or would I need to Bootstrap the mean, its
standard error and its confidence inetrvals and thus presenting a few
additional statistics over and above the means and its 95% CI?
Thank you very much indeed for taking the time to consider my
request. I look forward to hearing from you.
Yours faithfully
Panos.
Panos Papanikolaou
*********************
Research Fellow
Nursing, Health and Social Care Research Centre
SONWS
UWCM
Heath Park
Cardiff
CF14 4XN
Great Britain
E-mail: [log in to unmask]
Direct line: +44 (0)29 2074 2197
Fax: +44 (0)29 2074 5521
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|