Biking vs. walking USDOT survey



Status
Not open for further replies.
B

Blake McCully

Guest
The US Census Bureau estimates the US population as 290,909,012 as of May 7, 2003. A population of
9,600 was surveyed of 16 yr old and over. Without excluding those under 16 from the estimated
population, only 0.003 % of the population (as of today not in 2002 when the survey was don) is
statistically insignificant.

Of course, as we all know, you can make statistics say anything you want. I'm not sure what the
point of the survey was? Any suggestions.

B "John Riley" <[email protected]> wrote in message
news:[email protected]...
> http://www.dot.gov/affairs/bts0703.htm
>
>
>
> --
> >--------------------------<
> Posted via cyclingforums.com http://www.cyclingforums.com
 
S

Seth Jayson

Guest
> population (as of today not in 2002 when the survey was don) is statistically insignificant.

You really need to reevaluate your understanding of how surveys and statistics work. The whole point
of statistics is to extrapolate based on carefully-selected sample groups.

Statistical method (simplified by nice software packages like SPSS) allow us to take our sample, and
our survey results, and to use varying statistical analyses (like ye' famous Chi Square) to
determine whether the given responses are indeed statistically significant.

The old "you can use statistics to argue anything" is a red herring, and is usually the cry of
people who are frustrated with the media's faulty reporting on statistics.

(In other words, the offending reporter or newspaper wil say something like, "60% of our responders
prefer a 40 year-old Chinese upright to a Bacchetta Strada" without ever telling you anything about
sample size, sample method, survey method, or statistical tests)

With a proper survey and sample as large as 9,600, properly selected, it would be very easy to
arrive at findings that are statistically significant for the entire country.
 
N

Nigey

Guest
Warning, OT.

Very true -unfortunately the media is not interested (nor does it think its audience is -and they're
probably right) in how statistics can be gathered.

However, I still believe (even taking into account all the factors you mention) you're forgetting
that sometimes the human element is all important. A recent example for me was in the state of
Massachusetts when the population voted to support a "clean" elections bill. Can't remember the
exact question wording but it was something along the lines of do you support funding election
candidates to prevent external interests influencing candidates. It passed. Then (in my opinion)
when the political system realized it was in danger, it put up another question -it was along the
lines of "Do you want your tax dollars funding politician's campaigns?". Not unsurprizingly, it
failed. Apologies here for not recalling exact wording, but it was along those lines and I think you
get the gist of my argument here. And even better, you can always preface questions with "leaders"
that are more likely to promote a result you want -"forcing" the respondent to appear consistent.

So sort of the same question, but two different results? Sometimes the wording and context of the
question are all important. Bottom line: the old "you can use statistics to argue anything" isn't
necessarily a red herring!

cheers

Nige

[email protected] (Seth Jayson) wrote in message
news:<[email protected]>...
> > population (as of today not in 2002 when the survey was don) is statistically insignificant.
>
> You really need to reevaluate your understanding of how surveys and statistics work. The whole
> point of statistics is to extrapolate based on carefully-selected sample groups.
>
> Statistical method (simplified by nice software packages like SPSS) allow us to take our sample,
> and our survey results, and to use varying statistical analyses (like ye' famous Chi Square) to
> determine whether the given responses are indeed statistically significant.
>
> The old "you can use statistics to argue anything" is a red herring, and is usually the cry of
> people who are frustrated with the media's faulty reporting on statistics.
>
> (In other words, the offending reporter or newspaper wil say something like, "60% of our
> responders prefer a 40 year-old Chinese upright to a Bacchetta Strada" without ever telling you
> anything about sample size, sample method, survey method, or statistical tests)
>
> With a proper survey and sample as large as 9,600, properly selected, it would be very easy to
> arrive at findings that are statistically significant for the entire country.
 
M

Michael Plog

Guest
Concerning the statistical significance of a sample. A random sample means that every member of the
population has an equal chance of being selected. Any sample is expected to vary from what would be
obtained if the entire population were measured. This potential for variation is expressed in two
concepts. First is the Confidence Level, second is the Estimated Error rate. Several formulae (the
plural of "formula" can also be written "formulas") exist to determine what the confidence level
and error rate of a sample. A search engine will provide you with some of the mathematics involved.
This is not really the place to go through something like n=((t/e)**2*p(1-p))/1+1/N(..... you see
the problem.

Anyway, for the example you cited from the Census Bureau: For the population of almost 300 million,
a sample of over 9 thousand will give a Confidence Level of 99% with a tolerance for error between
.01 and .02.

Now, for the second thing that tends to bother professional statisticians and evaluators. Statistics
will NOT say anything you want. An unscrupulous person can find or distort some number that supports
an unreasonable position. The inaccuracies come from the misuse of statistics, not the mathematical
implementation.

Sorry to get on a high horse here, and I have not seen the field notes of the study. It is possible
a malicious person biased the selection process or the wording of the questionnaire. However, the
results that nearly three-quarters of the population wanted improvements for walking and cycling
seem impressive to me. The idea that as people age they cycle less seems to fit my observational
generalizations.

One lesson from the survey may be to promote recumbents more aggressively among older people.
Another lesson might be the encouragement of accessories designed for recumbent (older) riders, such
as helmets with wider brims and clothing with side (not back) pockets.

Enough said for now. I need to get a ride in before the rain hits.

Thanks, Michael Plog

"Blake McCully" <[email protected]> wrote in message news:[email protected]...
> The US Census Bureau estimates the US population as 290,909,012 as
of May 7,
> 2003. A population of 9,600 was surveyed of 16 yr old and over.
Without
> excluding those under 16 from the estimated population, only 0.003 %
of the
> population (as of today not in 2002 when the survey was don) is statistically insignificant.
>
> Of course, as we all know, you can make statistics say anything you
want.
> I'm not sure what the point of the survey was? Any suggestions.
>
> B "John Riley" <[email protected]> wrote in message
> news:[email protected]...
> > http://www.dot.gov/affairs/bts0703.htm
> >
> >
> >
> > --
> > >--------------------------<
> > Posted via cyclingforums.com http://www.cyclingforums.com
 
D

Doug Huffman

Guest
Almost all on the 'heat' here - the flames - come from naive use or understanding of statistics,
either formal or rhetorical. Thanks for a good post.

"Michael Plog" <[email protected]> wrote in message news:[email protected]...
> Concerning the statistical significance of a sample. A random sample means that every member of
> the population has an equal chance of being selected. Any sample is expected to vary from what
> would be obtained if the entire population were measured. This potential for variation is
> expressed in two concepts. First is the Confidence Level, second is the Estimated Error rate.
> Several formulae (the plural of "formula" can also be written "formulas") exist to determine what
> the confidence level and error rate of a sample. A search engine will provide you with some of the
> mathematics involved. This is not really the place to go through something like
> n=((t/e)**2*p(1-p))/1+1/N(..... you see the problem.
>
> Anyway, for the example you cited from the Census Bureau: For the population of almost 300
> million, a sample of over 9 thousand will give a Confidence Level of 99% with a tolerance for
> error between .01 and .02.
>
> Now, for the second thing that tends to bother professional statisticians and evaluators.
> Statistics will NOT say anything you want. An unscrupulous person can find or distort some number
> that supports an unreasonable position. The inaccuracies come from the misuse of statistics, not
> the mathematical implementation.
>
> Sorry to get on a high horse here, and I have not seen the field notes of the study. It is
> possible a malicious person biased the selection process or the wording of the questionnaire.
> However, the results that nearly three-quarters of the population wanted improvements for walking
> and cycling seem impressive to me. The idea that as people age they cycle less seems to fit my
> observational generalizations.
>
> One lesson from the survey may be to promote recumbents more aggressively among older people.
> Another lesson might be the encouragement of accessories designed for recumbent (older) riders,
> such as helmets with wider brims and clothing with side (not back) pockets.
>
> Enough said for now. I need to get a ride in before the rain hits.
>
> Thanks, Michael Plog
>
>
>
>
>
> "Blake McCully" <[email protected]> wrote in message news:[email protected]...
> > The US Census Bureau estimates the US population as 290,909,012 as
> of May 7,
> > 2003. A population of 9,600 was surveyed of 16 yr old and over.
> Without
> > excluding those under 16 from the estimated population, only 0.003 %
> of the
> > population (as of today not in 2002 when the survey was don) is statistically insignificant.
> >
> > Of course, as we all know, you can make statistics say anything you
> want.
> > I'm not sure what the point of the survey was? Any suggestions.
> >
> > B "John Riley" <[email protected]> wrote in message
> > news:[email protected]...
> > > http://www.dot.gov/affairs/bts0703.htm
> > >
> > >
> > >
> > > --
> > > >--------------------------<
> > > Posted via cyclingforums.com http://www.cyclingforums.com
> >
>
 
D

Dennis Pedrick

Guest
> So sort of the same question, but two different results? Sometimes the wording and context of the
> question are all important.

Here's the actual question from the survey. Looks pretty straight forward:

48c. (If respondent answered "Yes" to the previous question..."Would you like to see any
changes...") What changes would you like to see made in your community? (Open ended and code)
(Allow three responses) 01 Other (list)
48 (cA)03 (Refused) 04 HOLD 05 HOLD 06 More bike trails 07 More bike paths 08 More bike lanes 09
Allow bikes on sidewalks 10 Don't allow bikes on sidewalks

Regards -- DP
 
Status
Not open for further replies.

Similar threads