Research Methods

[sam0044]

Can some please put into simple words on what stratified sampling is and what random stratified sampling is?

I am confused with them

[Damoz.G]

Stratified Sampling is where the population is divided into subgroups (Based on similar characteristics such as age, sex, residential area, etc.), and then selecting a sample from each subgroup. So, its bias.

Random Stratified Sampling is where a sample is RANDOMLY selected from each subgroup.

Does that help at all?

[sam0044]

yeah it does, but other definitions say this: Involves breaking the population into distinct subgroups, or strata, then selecting a separate sample from each stratum, as the same proportions they occur in the target population.

What would this part mean???:then selecting a separate sample from each stratum, as the same proportions they occur in the target population.

[alondouek]

Okay, to give you an example:

I have a population that I'm trying to derive a sample from. I break this population up into strata based on a characteristic, for example 'skin colour'. Let's say that White skinned people make up a proportion of 0.7 in the population, while brown skinned people make up 0.23 and dark skinned 0.07 respectively.

When I take samples from the strata, the sample size for the 'White skinned' strata will be larger as I am trying to account for all random variation within the strata.

Does that help?

[sam0044]

yeah it does abit. I am not going to stress about it too much. I mean it would probably just be an MCQ

[Damoz.G]

Yeah, plus Random Stratified Sampling isn't commonly used because it is time consuming.

[adrihepburn]

Stratified Sampling is where the population is divided into subgroups (Based on similar characteristics such as age, sex, residential area, etc.), and then selecting a sample from each subgroup. So, its bias.

Random Stratified Sampling is where a sample is RANDOMLY selected from each subgroup.

Does that help at all?

What if the population of interest is very large (as it usually is) ? For example, 'children' - how can you divide every single child into subgroups?? isnt this very time consuming and virtually impossible? does this method of sampling ensure that the sample drawn from it is representative of the population?

[alondouek]

What if the population of interest is very large (as it usually is) ? For example, 'children' - how can you divide every single child into subgroups?? isnt this very time consuming and virtually impossible? does this method of sampling ensure that the sample drawn from it is representative of the population?

You probably wouldn't stratify based on 'children' alone, because that's impractical and nowhere near specific enough. Depend on your IV, you'd take your population as 'children', then stratify based on another characteristic.