A Summary of Sampling
After hours developing your research plan, writing your questions, and programming your survey, it is time to launch your survey. Before launching the study, it is essential to have a precise sample plan. This is a key step when fielding quantitative research. Proper sampling is important for making sure your research captures an accurate audience that your research is meant to represent. Here is a quick lesson on sampling, its importance, and how to properly plan your sample.
What is Sample?
You have an audience you want to reach that is unique to your objectives. It may be the US population, students, buyers of gum, or even physicians. Regardless of who your audience is, it is not efficient to survey the whole population due to several factors, including price and logistics.
That is where sampling comes in; it allows you to project the outcomes of a whole population based on a small subset. The larger the sample size, the smaller variance in error to match the population accurately. At a certain point, the variance becomes so small that increasing your sample size is no longer beneficial.
You will need to take the following attributes into account to help determine the sample size:
- Size of your overall population
- The margin of error (ideally, you want it to be less than 4%)
- The degree of uncertainty
- Confidence level (typically, you base it on a 95% confidence level),
- The percentage of certainty that the confidence interval would contain the true population parameter when you draw a random sample.
When considering a sample size, you must understand whether you are sub-segmenting the audience or conducting advanced analytics. Both tactics may require more sample.
If your sample is supposed to match a particular audience in characteristics, then it is essential to make sure it is representative. This means proportionally matching characteristics such as demographic, purchase habits or behavioral attributes. For example, if your target population skews slightly more female, so should your sample.
Making your sample representative is important because it helps ensure that your data truly reflects your wider target audience. People from various backgrounds and demographics will have vastly different opinions. Therefore, making sure your sample includes the proper proportion of each of these individual differences helps keep your data accurate.
How Do You Ensure Your Sample is Representative?
Proper representation is important, but how do you achieve it? There are a few ways to collect a representative sample, each with advantages and barriers. Choosing one of these methodologies can help prevent sampling error.
Stratified sampling involves specifying the needed cohorts within your sample and setting quotas for each. While specifically fielding for specific quotas, randomly selecting participants within these groups ensures unbiased data while still matching the sample to the population. The methodology provides high accuracy, but fielding the remaining open slots can become difficult and time-consuming as quotas fill.
Weighting achieves sample representation after survey fielding has already closed. The methodology involves adjusting the power of each respondent to make the influence of each sub-group representative of that population, even if the physical number of people in the sample is slightly off.
Here is an example. Let’s say you have too few responses in your sample from people under 30 years old and too many people above 50. Weighting can make the answers from the younger group “worth more” than the answers from the older cohort. This evens the playing field when looking at the data.
Weighting can be an incredibly useful tool, allowing you to match the population even after responses have been collected. However, it is important to use the methodology sparingly, as over-weighting could provide too much power to an individual respondent, degrading the quality of your results.
Combining stratified sampling with weighting
Using both methodologies to achieve proper representation can minimize the difficulties or challenges of using only one. Beginning with a stratified sample allows you to fill your survey with as much physical representation as possible. But when quotas start to fill up, and targeting becomes too narrow, minimal weighting allows you to stay on schedule by opening up to more general responses and making adjustments to correct the representation later on.
A One More Tip Before You Go…Use Databases
Often it can be difficult to pinpoint the demographic breaks for your sample. In these cases, there are several resources that you can turn to. The largest of these databases is the US Census. The Census collects countless data about the US population, from demographics to employment to homeownership. It is also free!
Using the census and other databases to proportionally plan your sample will help ensure your research adequately reflects the larger population. Other databases, such as MRI, can provide attributes in regards to behavioral characteristics. However, many of the courses have a price tag.
A well-planned sample will elevate your research and provide assurance that your analysis will be on-point, accurate, and statistically sound. If you don’t take the time to plan out your sample, you run the risk of degrading the quality of your data. It is crucial to take the time to gather the sample that fits your needs in order to set your analysis up for success.