Disproportionate Stratified Sampling

Question

Hi!

I have an undergraduate mixed-method thesis and my sampling technique is disproportionate stratified sampling technique. My strata are students and faculty members. The population of students is 1200 while faculty members is 30. My study is only limited to my department.

Now, I will be deploying a survey questionnaire with Part 1 and Part 2.

1. Part 1 - Multiple choice questions and dichotomous questions
2. Part 2 - Likert scale questions

My question: How do I compute for the sample size in each strata? I've been seeing some journals that use disproportionate sampling. I can't see how they compute, but I may be guessing that they just do it by "availability". For example, their total sample of the their strata is 385. Then they just administer the survey to their strata and whatever number they received the accomplished strata is the sample size of their strata. Is this how we do the disproportionate sampling? Or is there a formula?

Another question: I've been reading journals with disproportionate allocation. They made their survey have survey weights. A smaller weight to the big strata and larger weight to the small strata. They made this so that the bigger strata won't overpower the smaller strata. Is this how we also do in my study? If so, how to determine the survey weight in my strata? Is there a formula?

If you have any references I can read (books, websites, etc.) regarding this, please do send.

Thank you for your help!!

Answer 1

Hi Jared!

That's a valid question, here is a brief guide below:

Sample Size Calculation:

1. Stratified Sampling Formula:

- For proportional stratified sampling: n_h = (N_h / N) * n

- For disproportionate stratified sampling, you can assign different sampling fractions to each stratum based on factors such as stratum size, variability, or importance.

2. Disproportionate Sampling by Availability:

- If you are allocating based on availability or practical constraints, you can indeed use the "availability" method. It involves surveying all individuals in the stratum who are available and willing to participate.

Survey Weights:

1. Purpose of Survey Weights:

- Survey weights are used to correct for unequal probabilities of selection and potential biases introduced by disproportionate sampling.

2. Weight Calculation:

- The weight W_h for each stratum can be calculated as the inverse of the sampling fraction: W_h = 1 / n_h.

3. Balancing Strata:

- Assign smaller weights to larger strata and larger weights to smaller strata to balance their influence in the overall analysis.

Symbols definition:

1. n_h​: Sample size in stratum h.
2. N_h​: Population size in stratum h.
3. N: Total population size.
4. n: Total sample size.

It's important to note that obtaining a more tailored answer would require a more in-depth examination of the specific details of your data and research objectives. While the general guidance provided offers a foundation, the nuances of your study, population characteristics, and research goals may warrant a closer look. For a more precise approach, consider consulting with an expert (👋) who can delve into the specifics of your data and provide tailored guidance.