>> Statistically speaking, both strongs are sensitive to even small errors in polling. Its not Rass fault, one out of every twenty polls is a dud with big samples. With small samples, its even more frequent. <<
Good explanation! Needs to be pointed out time and again, since so many readers of the daily “Rasmussen thread” here on FR have no background in statistics.
>> The three-day rolling average is supposed to iron them out <<
Yeah, but I think “supposed” is the critical word:
We’ve seen time and again that unfortunately, just one night’s “severely out of range” poll results can contaminate Ras’s reports for three days’ running.
In this regard for example, last week there were three consecutive days’ worth of “Index” values that seemed too low for any logical explanation, followed by three days of numbers that were way too high. Therefore, I believe a ten-day moving average (or something like it) is a much more reliable indicator of the true state of public opinion.
Today, the ten-day average stands at -15.4. That’s higher than I had been expecting, since last week it got down to -17. At this point, it’s probably best to say only that the “true, underlying” value of the Rasmussen Index is in the range of -15 to -17. And in my book, such a value is just plain excellent!
By the way, speaking of averages, we can now see that the average for the month of May was -13.5, which is a marked improvement over the average value of -10.9 we suffered for the month of April. So the good news continues to roll in!
Thanks for the applied science. :)