We're doing something a little different than usual. Over the past year, we've noticed our clients are a curious bunch, which is awesome.
Identifying patterns is how we make a living, so compiling the most frequently asked questions was the most appropriate choice.
How is Sample Size Determined?
Sample size is determined by a few key factors. But first, a quick refresher on sampling:
- Sampling is taking a small subset of a group of people and estimating the characteristics of the entire group based on the characteristics of the subset.
Now that's out of the way, determining sample size comes down to maintaining statistical reliability/significance, while also factoring in client preferences. The three main factors that contribute to sample size are:
- Population size
- Confidence level
- Acceptable margin for error
All three of those factors have a direct impact on each other, and therefore a direct impact on sample size. The industry standard for statistical significance is a minimum +/- 5% margin for error at a 95% confidence level, so it really comes down to population size and how difficult it is to reach the target audience. Statistically speaking, sample size doesn't change much once a population reaches 20,000 members, but the complexity of the target audience can have a major influence on sample size.
- For example, there's about 200,000 CEOs in the United States, but they're much more difficult to reach than, say, homeowners.
While both populations certainly have more than 20,000 members, anyone who has a boss knows that CEOs are much more difficult to get in contact with, therefore their sample size is smaller compared to homeowners.
What Happens if an Incomplete Survey is Received?
Missing data values are the bane of data analysts everywhere. Yet, like death and taxes, they're inevitable. While many data scientists and analysts typically impute missing data values, that doesn't really fly in the world of market research.
Market research studies are opinion-based, with each respondent giving their own unique insight. When a value is imputed, the algorithm essentially predicts the missing response based on the frequency of responses in the dataset as a whole.
- For example, if 150/200 people answer "yes" to a question and 49 people answer "no", the algorithm will use those 150 "yeses" and 49 "nos" to try to predict the missing response.
We're not big into letting algorithms predict people's opinions for them. Typically, we'll either treat the missing value as a refusal to answer the question, or, if the client prefers, we'll scrub the incomplete survey from the dataset.
How is Data Analyzed?
We've covered this in-depth in a previous edition, but here's a quick refresher. Data is collected and compiled into a single location such as an Excel spreadsheet or SPSS file. Our data analysis team has many different methods of analyzing collected data in their toolbox, but the most common ones are:
- Multiple Regression
- Factor Analysis
- Cluster Analysis
- Multidimensional Scaling
If you're craving a deeper dive into the world of data analysis, you can find more here.