If you haven’t read the first part of this two-part series, which was published before the primary results were known, please go here to review it.
The final results in the New Hampshire Republican presidential primaries have been published and I’m confused.
Based upon the sample from Dixville Notch, each candidate should have received one portion of the votes, but when a larger sample size was taken, the candidates all received different portions of the vote. None of the final results were even close to the predicted results from Dixville Notch.
Jon Huntsman’s scores were off by 31%
Ron Paul’s scores were off by more than 100%!
Rick Perry’s scores were pretty close, but he still got votes even though the survey said that he wasn’t supposed to get any votes at all.
Rick Santorum should be happy. The initial survey said that he would get 0.000% of the vote. He ended up with 9.4% of the vote. What a massive error that was!
|Candidate||Dixville Notch Predicted Percent Vote||Actual Percent Vote||Difference|
|Newt Gingrich||1/9 = 11.1% of vote||9.4% of vote||- 18%|
|Jon Huntsman||2/9 = 22.2% of vote||16.9% of vote||- 31.3%|
|Ron Paul||1/9 = 11.1% of vote||22.9% of vote||+ 106%|
|Rick Perry||0/9 = 0% of vote||0.7% of vote||+ “a lot” %|
|Mitt Romney||2/9 = 22.2% of vote||39.3% of vote||+ 77%|
|Rick Santorum||0/9 = 0% of vote||9.4% of vote||+ “even more” %|
Now you can see from the sample of an actual Press Ganey report in the previous post, a difference in a “mean score” of less than 1% can make the “percentile rank” change by 5 points or more. What happens when the mean score changes by 100% or more? Disaster.
So it seems that the scores from small sample sizes in Press Ganey surveys are pretty much like random number generators.
Kind of like administrators praising and demeaning staff based upon how many blue Chevrolets drive past the hospital between 3PM and 4:30PM on Tuesday afternoon. If there are 5 Chevys, you’re the best. If there are 0 Chevys, you’re fired.
If administrators and hospital boards can’t see the problems with basic statistics, how are they supposed to run hospitals effectively?
Make sure to leave a copy of this post on your administrator’s desk or under your administrator’s door.
We need to bring some accountability to those who manage our profession.