When the pancakes are ready the couple sits down at the kitchen table. They excitedly bite into the pancakes. After swallowing a few pieces they cringe with disgust and look at each other in astonishment. "Why on earth do rich people like to eat potato pancakes?" they ask each other.
Translate that story into preparing political polls (or cooking them). You don't have enough African Americans in the sample compared to their representation in the general population you are polling? No problem. Just replace that omission with a number using a mathematical formula. Too few millennials? Boost their representation with a number. Not enough elderly--just assign a number to compensate. The underlying assumption is that these mathematical weightings are equivalent to actually polling more human subjects equal to their percentage in the overall population being polled.
Is that assumption valid? We can't be sure.
The researchers believe that assigning greater values to small numbers of participants in underrepresented samples--weighting--is almost as good as polling more human subjects. This enables them to announce the "validity" of their findings. But depending on the number of weightings, what they may be offering for public consumption is the Russian peasant potato pancakes.
Bon Appetite!
Note:
While we can't trust the findings of a poll with a small non-representative sample if there are a large number of such polls, collectively they may represent a significant segment of the population being studied. Currently, a plethora of individually flawed polls due to small samples all give Biden a wide margin over Trump. Thus, it is reasonable to speculate that Biden probably does have the lead. Keep in mind though that most often there are only singular or a small number of polls for many elections and polls and surveys on other topics--e.g. how many prefer this or that. If the sampling in these polls is inadequate we can only reject or remain skeptical about the reported results.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
1
1
1
