The condfidence intervals for the counts are so tight that
the dot for the mean overplots almost all of the them.

This has given several students the impression that the
changes they made to draw the intervals were not correct,
when in fact there were correct.   


To avoid this problem in the future I have modified the assign
to use 99 percent confidence intervals.  

conf90 = qnorm(.95) = 1.644854
conf99 = qnorm(.995) =  2.575829

This years students that have not already posted their
assigns are free to use a large value for conf. 


This example is related to the problem cited in class.
The typical confidence interval encoding draws the most
visual attention to the least accurate estimates. 

Here, since the bounds are hidden, the assumption is likely
to be that nothing is known about the quality of the estimates.
This is really wrong.  


Do you have suggestions?  
Should we plot gold stars for really good estimates?