You are viewing a single comment's thread. View all
0
devotech2 on scored.co
1 year ago0 points(+0/-0)1 child
You can get a reasonable, generalized average based off of a much, much smaller amount of people than an actual large population, as long as the subjects are wide ranging themselves, given that overall trends can be observed with a small group of people (example: "everyone at my high school started dressing up in corduroy pants and cashmere sweaters. It must be a fashion trend among teenagers in general now", and this would probably be reasonably accurate). The number alone is not inherently disingenuous. The +/- just seems pedantic.
But there's a problem when it concerns bias of the people making the polls targeting specific demographics (which with a lump sum of people such as "the American populace" is incredibly easy to do), and also the demographics of those more likely to take part in polling vs not in the first place. Or just making shit up, which is not unlikely at all. I can be a person who hates pizza and just make up some bullshit graph on Google charts "proving" my statement that everyone hates pizza, and say I interviewed x number of people, and nobody would be able to reasonably tell me I did not do so.
But it doesn't matter anyways, conservatives are as useless as liberals. Perhaps even worse. I don't even care if 90% of Americans legitimately want Kamala Harris to win. Trumpers are almost entirely a lost cause if they haven't put 2 and 2 together at this point. Who gives a fuck?
Ayh want to argue your small model theory for a few reasons
1. As you sort of mention, bias is key. Issue is that in our modern society, bias is built in to city vs rural and cities are the primary driver of polling data collection.
2. They do not ask point blank in small model data collection. They ask to rank perspective. "How do you feel about the state of xyz?" And they can do this bcz small model collection and thus flex data to suit, thus also why the 3% "error"
3. No. Just no. Conditioning tolerance of a small model is the same issue in sciences. Inference is not a good nor viable methodology to knowledge. It is pure propaganda.
But there's a problem when it concerns bias of the people making the polls targeting specific demographics (which with a lump sum of people such as "the American populace" is incredibly easy to do), and also the demographics of those more likely to take part in polling vs not in the first place. Or just making shit up, which is not unlikely at all. I can be a person who hates pizza and just make up some bullshit graph on Google charts "proving" my statement that everyone hates pizza, and say I interviewed x number of people, and nobody would be able to reasonably tell me I did not do so.
But it doesn't matter anyways, conservatives are as useless as liberals. Perhaps even worse. I don't even care if 90% of Americans legitimately want Kamala Harris to win. Trumpers are almost entirely a lost cause if they haven't put 2 and 2 together at this point. Who gives a fuck?
1. As you sort of mention, bias is key. Issue is that in our modern society, bias is built in to city vs rural and cities are the primary driver of polling data collection.
2. They do not ask point blank in small model data collection. They ask to rank perspective. "How do you feel about the state of xyz?" And they can do this bcz small model collection and thus flex data to suit, thus also why the 3% "error"
3. No. Just no. Conditioning tolerance of a small model is the same issue in sciences. Inference is not a good nor viable methodology to knowledge. It is pure propaganda.