Originally Posted by
spare_wheel
The spike to ~6% in PDX was well outside the margin of error.
[TABLE]
[TR]
[TD="align: right"]
2008:| 17365[/TD]
[TD="align: right"]2061[/TD]
[TD="align: right"]5.96%[/TD]
[TD="align: left"]
2007:|[/TD]
[TD="align: right"]10987[/TD]
[TD="align: right"]1587[/TD]
[TD="align: right"]3.91%[/TD]
[/TR]
[/TABLE]
(counts, margin of error, percentage)
https://docs.google.com/spreadsheet/...=sharing#gid=0
Originally Posted by
mr_bill
...
Now go back at the OP and look at the Portland ACS data and add in the margin of error. Again, all you can really say is there are more primary bicycle commuters in Portland since 2008 than 2005. [EMPHASIS ADDED] However, the change of commuters since 2008 is small compared to the margin of error, which is large! We can't say if it is flat, increasing, or decreasing. Panic over not knowing from the ACS? Or use other, more precise numbers that Portland collects?
The Boston ACS numbers are even more all over the place, and other than there are more cyclists commuting primarily by bicycle than in 2005, there's not much more to say. But local measurements tell the story.
That's my only point.
...
-mr. bill
Please be careful about switching back and forth between mode share and counts, especially with the ACS-1 year numbers. You have a noisy numerator and a noisy denominator. If you want to talk mode share, talk mode share.
This is really not hard stuff. For example, to make it clear what error you are repeating - the Boston ACS-1 year numbers for:
2007: 1.0% +/-0.3%
2008: 1.6% +/- 0.4%
The delta between 2008 and 2007 is 0.6%, outside of the margin of error, right? Wrong.
The 2007 confidence interval is between 0.7%...
1.3%
The 2008 confidence interval is between
1.2%...2.0%
The two noisy values overlap. You see a spike that may or may not be there - you simply can't tell from the ACS 1-year numbers. And you refuse to look at any other numbers. So I don't know what to tell you, other than repeat again. From the ACS-1 year data, Portland since 2008 has more bicycle commuters (>=5% share) than they did before 2008 (<=5% share). There is nothing else to see there.
You are making the same mistakes that so many minor party candidates make with polls (and astonishingly, one MAJOR party candidate who really didn't think he was going to lose made).
1- Small numbers changing a small amount but within the margin of error shows a HUGE swing to my candidate.
Wohoo, my candidate's numbers went from 5%+/-3% to 10%+/-5%. Look at that spike - we doubled!
2 - Finding flaws in *EVERY* method of measuring that doesn't tell the story you believe is happening. Straight FUD.
Wohoo, my candidate has momentum, there are inherent "biases" in the pools that show the incumbent 5% ahead when really, after adjusting for the inherent "bias" or ignoring the "biased" polls, our guy is blowing him away. And besides, Tioga county!
-mr. bill