So, when doing polling, we use binomial distribution to capture the data (in case there are only two choices.). My question is, let us say that estimated percentage according to the poll that supported Obama was 47%. Then we would be able to set $p = 0.47$ and calculate standard error when given the number of samples, $n$. But the true standard error will not be this. So, is it OK to use estimated standard error right away?
Asked
Active
Viewed 25 times
0
-
"use estimated se"... for what purpose? Without some context about goals, your question cannot be answered. – Harvey Motulsky May 08 '13 at 23:58
-
@HarveyMotulsky For polling as in example. (see Wikipedia "margin of error" and how they gave an example of Kery - my question is, standard error in that article seems to have been derived from $p$ and $n$ - but that's not really true standard error. So, is it OK to use the estimated one?) – Poll May 09 '13 at 00:05