Different sample methods in statistics

[i_love_science]

What is the difference between stratified and quota sampling of a population?

For example, you can choose 200 males and 200 females from a state by quota sampling; or collect raw data first, stratify it, and then choose 200 males from one subsample and 200 females from the other subsample.

Is whether the data is sampled first or collected first the only difference between the two types of sampling?

Thanks.

 

[member 587159]

The stratification might be problematic. For example, when you have not enough samples of either males/females.

 

[i_love_science]

The stratification might be problematic. For example, when you have not enough samples of either males/females.

But doesn't that not matter? The raw data first collected for stratified sampling is from the entire population of the state, which is the same population to choose from for quota sampling? So if you don't have enough samples of either males/females, then it might be problematic for both techniques? Please correct me if I am wrong.

Are there are any other differences? Thanks!

 

[Office_Shredder]

The main difference in my mind is that quota sampling doesn't tell you how to sample. You get 1,000 potential subjects - if you are stratify sampling you pick 200 males randomly and 200 females randomly. If you are quota sampling, what do you do? By definition it's whatever you want.

Even if you are not trying to be malicious about this, you can get very bad results. If you take, say, the first 200 males and females to sign up, you might get people who are particularly interested in participating in your study, or only people who wake up early in the morning. If you take the first 200 alphabetically you might get mostly people from cultures that have lots of names starting with a.

On the other hand, quota sampling has the advantage of not requiring you get 1,000 people in the first place. If you're going to just take the first 200 of each gender that sign up you can finish your sign ups much faster. In certain examples it might even be superior to use quota sampling - if you're doing a study on whether men or women are better able to survive a mission to Mars by putting them through stress tests, and you want to pick 5 of each, picking randomly from among all your volunteers will be subpar compared to looking at their skills and fitness levels and picking the five best candidates.

 

[Stephen Tashi]

What is the difference between stratified and quota sampling of a population?

That's a good question. Stratified sampling has a mathematical definition. Descriptions of quota sampling on the web say that it differs from stratified sampling by being non-probabilistic. Probability theory (and hence mathematical statistics ) doesn't have much to say about non-probabilistic sampling!


A sampling method may have stochastic aspects but still be considered non-probabilistic with respect to the population being analyzed. Your example doesn't make it clear what population is being analyzed or how you select the raw data. If the population being analyzed is (only) the population of the state and the raw data is selected in a probabilistic manner from that entire population then what you describe may not be quota sampling.

 

[pbuk]

collect raw data first, stratify it, and then choose 200 males from one subsample and 200 females from the other subsample.

Is whether the data is sampled first or collected first the only difference between the two types of sampling?

That is not what stratified sampling is, you do not collect the data first.

In stratified sampling you divide the population into sub-populations and then take unbiased random samples from each sub-population.

As implied by @Office_Shredder and @Stephen Tashi above, in quota sampling you divide the population into sub-populations and then take (possibly biased) samples from each sub-population.