At 04:14 PM 12/3/02 -0800, Jill Binker wrote:
did they select 1221 (and why THAT number?) and find each and every one of them? no way i can believe that ... given that many are not home ... or won't be during that narrow window of calling ... or won't talk to them even if they answer the phone ... or, they find no eligible voter in the house ...They call back again, and again, and again (actually, when they first reach you, if there are more than 1registered voters, they have random methods for asking for 1 specific person, and call back if they aren't there). Part of the reason for keeping the sample as small as possible is so they have the resources to do the follow-up to actually reach the 1221 people.
this is not much different that surveys where .. you get a first mailing ... send a second one when you only got 34% the first time ... do it a third time ... you are still VERY LUCKY if you get 75% ... why should we believe that using the phone will be much different ... even if they keep calling back?
with valiant efforts if we START with 1221 ... we will not wind up with 1221 ... no where close
i contend that ... the samples that are taken this way ... cannot be considered to be SRSs ... from the likely voter population ... they mayBE considered fairly close to SRS of phone subscribers who happen to be home when you call AND, are willing to talk to you about politics ...
does anyone know offhand ... out of perhaps 1000 names you call:
1. what percent you canNOT get in touch with?
2. and, in addition, how many you do get in touch with DON'T meet one or more of the criteria you have set for discussing the issues(s) and
3. in addition, what % just flat out won't discuss the matter with you EVEN if they meet your criteria
we can go in the opposite direction ... and say that we want to end up with 1000 of that ilk but, if so, we will be starting with maybe 2000? 3000? NOT 1000
finally, anyone can take a sample by any method and by luck ... have it be as representative of the target population as one done precisely via SRS ... BUT, the fallacy in that "method" is that ... you were just flat out lucky ...
it is only when you know that you have used a SRS method ... that you can argue the "trust" factor in your results because, that is the only way you can estimate some statistical margin of error
it is not n (though it plays a role) ... it is how you got the sample that counts
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
