My apologies... I was in a class of four people in the sixth form that did a Statistics A-Level....
On 26/09/2007, vijay chopra <[EMAIL PROTECTED]> wrote: > > > > On 26/09/2007, Christopher Woods <[EMAIL PROTECTED]> wrote: > > > > Leaving the last digit from the last octet out would be fine, though? > > Then you could group by IP addresses for purposes like fraud checking and > > suchlike. I'm sure the BBC sites always say that standard information such > > as browser and IP address will be collected whenever you submit information > > to the server, so that's a fairly standard get-out clause. > > > > That's actually a really good idea, and to add to my previous email, it > would certainly be intresting to see what topics inspire the most vote > fraud. Having Geographic and ISP info aswell would be good. Are Northerners > or Southerners more honest online? NTL customers or BT customers etc. > What utter tosh. I'm sorry, but aside from the fact that you cannot determine anything at all from an IP address, because of NAT and corporate gateways and proxy servers, firewalls and so forth, it misses out the principles of: - Psephology - IP addresses might not be pebbles, but you need to understand the actual system you are considering and not make generalisations about questions not yet even asked. - Statistical weighting. Unless you do a "universal poll", you should weight the incoming votes you get so that they represent the population as a whole. So, if you ask people to vote, and 25% of the voters are men, you need to weight the male votes up so they match the 50% in the population as a whole and "unweight" the female votes from 75% to 50%. - Secret ballots. The whole point of a "secret ballot" is that you do not know the votes of other people and cannot be influenced by votes already cast. This is not the case with most web, radio and telly voting where you are being encouraged to part with money, not provide a statistically correct outcome. Remember that ALL the voting where you are asked to pay for the call or text are simply revenue collection systems, not statistically valid ones. The "adjudicators" (on Big Brother for example) simple verify that the number of calls have been made, not the meaning of the votes. As far as I know the only systems that the BBC uses on a regular basis that are statistically valid is the popular music chart and the BARB figures. There's bugger all you can really do with an IP address, even a complete > > one, unless you're a malicious fellow with a botnet behind you. > > > > I know that, you know that and everyone on this list knows that, but it > doesn't make as good a headline in the daily mail as "BBC giving out > information about your computers" or "BBC helps spammers" then going on to > detail all the evil things that can be done with a botnet... > Or am I being too cynical? There is no reason why the BBC could not use a table where random values are assigned to each IP address as they are encountered - as long as a reverse look up was not published Vijay. > -- Brian Butterworth www.ukfree.tv
