How do I interpret Margin of Error?
What is Margin of Error (MOE)?
Margin of Error, or MOE, is a statistical measure that reflects the amount of random sampling error in survey results. A bigger MOE for a particular datapoint means a lower level of confidence in that value.
MOEs are usually represented as ±X%. This means if a datapoint is 10% with an MOE of ±2%, the true value of the datapoint is somewhere between 8-12%.
How MOEs are calculated in Latana
When you’re calculating MOEs for traditional survey data, you typically use a formula based on sample size, standard deviation, and confidence level.
At Latana, our approach to MOEs is based on how we use our MRP model to generate estimates. With MRP (Multi-level regression and poststratification) we are using data from the entire population to improve the quality of data, especially for hard-to-reach groups. The MRP model generates 100 estimates of each KPI for each slice of the population, and then we use the mean of those estimates as the value for the KPI.
For example, the MRP model generates 100 estimates of Aided Brand Awareness for high-income, high-education, urban, millennial females. The mean of those estimates is 45%, and that’s what we show as the value for Aided Brand Awareness for that segment. The lowest value of the 100 estimates is 42% and the highest value is 48%, which means it has an MOE of ±3%.
How to use MOEs in Latana
For waves that include MOEs, you have the option to turn them on and see them in the dashboard. They show as error bars on the charts, and green/yellow/red indicators to indicate high/medium/low levels of certainty.
You can use the MOEs to determine whether or not a change is significant. For example:
If Aided Awareness is 65% ± 3% in wave 1 and 77% ± 2% in wave 2, the change is significant and you can be confident it reflects a real-world change.
If Brand Consideration is 35% ± 3% in wave 1 and 33% ± 2% in wave 2, the change is not significant and it likely does not reflect a real-world change.