Re: Statistics... in http://www.dslreports.com/forum/r18466528 en Sat, 28 Nov 2009 19:23:15 EDT Sat, 28 Nov 2009 19:23:15 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18469440 Fluker : I totally agree.

On campus, the greater majority of people have a cell phone and broadband, but no landline.

I can't think of any student I know with a real phone. All of them have at least cable internet though. I doubt any of these people could have been included in this statistic if it were call based.]]> http://www.dslreports.com/forum/remark,18469440 Thu, 07 Jun 2007 23:24:56 EDT Re: Don't worry... http://www.dslreports.com/forum/remark,18469381 BIGMIKE : 53% Of US Homes Broadband Connected


And here comes the taxes!
Internet Taxes to Arrive by This Fall?
»www.trendplex.com/editorials/int···his-fall
--
Type "miserable failure" in Google]]> http://www.dslreports.com/forum/remark,18469381 Thu, 07 Jun 2007 23:12:07 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18468005 tschmidt :
said by Jerm See Profile :


I'm not sure why you think it would be flawed. I mean how much better can you get than a random sample? They calculated to margin of error as +/- 2.5% so that should give us some sense of accuracy.
I would agree if it were in fact random but it is not.

They called folks about broadband. That means only people with Wireline phones were included in the survey.

Did they randomly select x number of people and call those numbers ignoring non respondents or did they create a pool of numbers and call until they reached the number they needed to be statistical accurate? Each approach creates different biases.

When did they call and did they call at the same local time in each time zone?

/Tom]]> http://www.dslreports.com/forum/remark,18468005 Thu, 07 Jun 2007 18:55:13 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18467955 McSummation :
said by Jerm See Profile :

Take all the phone books in the US ...
Then that may automatically eliminate all the folks that don't have landlines, but have only cell phones.]]> http://www.dslreports.com/forum/remark,18467955 Thu, 07 Jun 2007 18:45:54 EDT Re: Don't worry... http://www.dslreports.com/forum/remark,18466840 RARPSL :
said by satellite68 See Profile :

quote:
The findings for the study, "Broadband Access and Service in the Home 2007," are based on a telephone survey of 1,600 randomly selected households across the country. The overall sample has a statistical margin of error of plus or minus 2.5%


Can anyone say flawed statistical methodology? How random? What percentage of those "randomly selected" households were in statisically-favorable-to-the-intended- outcome areas?
How many of these 1600 have non-broadband Internet (AKA Dial-Up) or have no Internet access?

IOW: Did they keep calling households until they get 1600 Internet Users and thus rejected 3200 households with no Internet (Thus meaning that Internet access was only 33% of the 4800 households actually called)?

Note: My 3200 non-Internet number is a manufactured number to show how the numbers could be slanted and does not mean that I claim/think the ratio of non-Internet to Internet households is that high.]]> http://www.dslreports.com/forum/remark,18466840 Thu, 07 Jun 2007 15:42:17 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18466550 satellite68 : Your assumption is incorrect; you are assuming that they followed your example-"randomly" calling 1600 households.
I doubt that happened here. This issue is highly politicized and often these kind of surveys are funded by (gasp) the major telecoms and/or policymakers or their brethren. I'm betting they "prequalified" the respondents to get the outcomes requested/demanded by whomever funded this survey.

It's easy to lie with statistics.]]> http://www.dslreports.com/forum/remark,18466550 Thu, 07 Jun 2007 14:53:46 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18466528 satellite68 :
said by Ahrenl See Profile :

Statistical Bias is very easy to manufacture, so I would first look at who provides their funding, and why the research was done in the first place. I'm not saying the numbers are incorrect.
Thank you, that's what I would've said sooner had you not beaten me to the punch. ]]> http://www.dslreports.com/forum/remark,18466528 Thu, 07 Jun 2007 14:49:36 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18466523 lesopp : How many people stuck on dial up couldn't participate because those collecting the data couldn't call them?]]> http://www.dslreports.com/forum/remark,18466523 Thu, 07 Jun 2007 14:49:19 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18466446 bmn :
said by Jerm See Profile :

I mean how much better can you get than a random sample?
A representative sample... Random samples are not always representative.
--
Prove it...]]> http://www.dslreports.com/forum/remark,18466446 Thu, 07 Jun 2007 14:38:07 EDT Re: Statistics... http://www.dslreports.com/forum/remark,18466349 Ahrenl : Statistical Bias is very easy to manufacture, so I would first look at who provides their funding, and why the research was done in the first place. I'm not saying the numbers are incorrect.

I do find it interesting that the under $50,000 households % of broadband users is growing faster than the over $50,000 households. Maybe they're only lower overall because they're generally less informed. ]]> http://www.dslreports.com/forum/remark,18466349 Thu, 07 Jun 2007 14:20:10 EDT Statistics... http://www.dslreports.com/forum/remark,18466021 Jerm : Thats the beauty of statistics, the results of a small sample are generally spot on when comparing against the "whole".

I'm not sure why you think it would be flawed. I mean how much better can you get than a random sample? They calculated to margin of error as +/- 2.5% so that should give us some sense of accuracy. I guess you'd have to know their exact method to know if it was flawed. Take all the phone books in the US and randomly select 1600 households via computer and call them up - doesn't seem too hard to me, and I'm not sure why you'd be wanting more accuracy than that.

There are a few arguments that could be made though. For example the people have to have phone service. But in general I think the statistic is probably correct and seems to fly in my book. Why do you think otherwise?]]> http://www.dslreports.com/forum/remark,18466021 Thu, 07 Jun 2007 13:11:35 EDT Re: Don't worry... http://www.dslreports.com/forum/remark,18465989 ColorBASIC : True; usage rates may be much much higher. :D]]> http://www.dslreports.com/forum/remark,18465989 Thu, 07 Jun 2007 13:04:23 EDT Re: Don't worry... http://www.dslreports.com/forum/remark,18465960 satellite68 :
quote:
The findings for the study, "Broadband Access and Service in the Home 2007," are based on a telephone survey of 1,600 randomly selected households across the country. The overall sample has a statistical margin of error of plus or minus 2.5%


Can anyone say flawed statistical methodology? How random? What percentage of those "randomly selected" households were in statisically-favorable-to-the-intended- outcome areas?]]> http://www.dslreports.com/forum/remark,18465960 Thu, 07 Jun 2007 12:58:06 EDT Don't worry... http://www.dslreports.com/forum/remark,18465866 ColorBASIC : These increases will still be spun to "US broadband sucks".]]> http://www.dslreports.com/forum/remark,18465866 Thu, 07 Jun 2007 12:38:54 EDT