You can report the uncertainty of your estimates as it is. If you plot the
histogram of the posterior distribution, that's all the information you
have about your quantity of interest. Also, you can calculate the
probability that there exists an positive effect on turnout.
Kosuke
On Mon, 12 May 2003, Anna Lorien Nelson wrote:
Hi all,
We have consistent patterns in our data that show ballot fatigue.
However, due to our very small data set sizes (115 to 138 pooled
time-series cross-sectional observations in each data subset) and the need
to calculate panel corrected standard errors, we are having difficulty
finding results that are statistically significant with 95% confidence.
Is it legitimate to report results with, say, 80% confidence? For
instance, when we drop to 80% confidence, we find that low (but
non-zero) number of initiatives on the ballot increases turnout an average
of 6.4% over elections where there are a fairly high number of initiatives
on the ballot. This seems like decent evidence of "ballot fatigue," and
the point estimate of a 6.4% difference in turnout seems quite politically
significant.
So, is this sort of result -- with 80% confidence -- legitimate to report?
Thanks,
Anna
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l