Suppose that you wish to know if a box of ball bearings is "up to standard". You could inspect the entire lot. But that's expensive. So inexpensive effective methods exist for drawing random samples from production lots for inspection. Similar methods exist for conducting a sample opinion poll to infer general opinion on certain topics. By such methods, what you learn about the sample may be inferred about the lot (or population) from which you take the sample.But why random? Because we need to avoid all of the kinds of biases we know of which could make the sample nonrepresentative. There are two true and famous stories about biased sampling -- one about polling, the other about sampling from events.
These stories show the need for sampling which avoids known sources of bias. Two of these stories show how nonrandom bias can arise. And the third story -- based upon a personal anecdote -- shows how stratum bias can arise in a sample rrandomly -- due to "sampling variability". Briefly, if gender is important to your enquiry -- that is, women may have different experiences or opinions than for men -- then the gender-stratum is critical for your survey or sample. To avoid one kind of bias, the proportion of this stratum in your sample should be the same as that in the population you're sampling from. (If 51% of the American population is female, your sample should show 51% female.)
So, if such "demographic" attributes are critical to your enquiry, you should use the demographics of the last census to tailor your sample-strata -- say, gender, race, age, education, etc. BUT THESE ARE QUOTAS!!!!! And they have been used thousands and thousands of times in the past and are being used every day -- by the Federal Government, by State and Municipal Governments, by private institutions for profit and nonprofit, etc. STRATIFIED (QUOTA) SAMPLING is the only way to avoid stratum bias UNLESS YOU SAMPLE NEARLY THE ENTIRE POPULATION! And this, in most cases, would not only be prohibitively expensive, but physically impossible!So, all sorts of important (but hidden) decisions, implementations of laws and regulations, plans, etc., which effect the employment and education of all of us DERIVE FROM QUOTA-SAMPLING! Yet opponents of "equal opportunity" legislation for women and minorities OBJECT TO DOING THE VERY THING THAT HAS BUILT THE STATUS QUO!!!!!
THIS IS THE DIRTY LITTLE -- DOUBLE-STANDARD! -- SECRET ABOUT QUOTAS!
I challenge any opposer of affirmative action to submit to a double blind test mattwer:two sets of data will be chosen:
I believe these results, if performed, will show that we live under "a double standard", not accepting significant statistical deviations on materials and service but allowing it in representing in this Republic women, childrne, and various minorities.one set is composed of measures or counts from a "universe" not involving affirmative action or demographic data, but possessing a mean value;
- one set consisting of demographic data in various categories showing difference from mean of its "universe".
- these data sets will be prepared by one group of men;
- another group of men -- without knowing the origin of the data -- will decide if the given value is "representative" or not.
.