Survey Data Imputation

The value of your research comes from the quality of your data. Survey questions are chosen because they allow you to run data analysis in order to draw conclusions. If there is a problem with your ability to collect data, you may not be able to run those analyses.

One issue – an issue that thankfully has been reduced over the years thanks to advancements in survey research technology and survey methodologies – is what happens when a data point is missing from the dataset. Those missing data points may be needed to run your analysis, and without them there could be problems with either the way you analyze data, or the programs you use to run the data.

Missing data may be due to a host of different reasons. Research dropout is one of the most common reasons, especially in longitudinal studies. Missed appointments can cause data to be lost, as can errors with data collection or a glitch in a database.

Sometimes the missing data is very small – as little as a single datum. Other times it can be considerable, like an entire respondent’s answers missing. When you need a completed dataset to run your analysis, you need a way to place that missing data back into the dataset.

You can do that with a technique called “Imputation.” Imputation is a way to place some type of value into the missing data points in order to have a complete database and run the analysis. Since the data is missing, the results will not be exact, but what the researcher can do is integrate imputation techniques to get the data as close as possible without knowing exactly what the data is. An example of such a technique would be where the researcher uses analysis tools to find other respondents that seem to be similar to the individual, based on the data that is available.

They would then compile all of the answers those individuals gave, and choose data at random from those individuals to fill in the dataset.

Imputation does reduce the accuracy of your data, so if analysis is possible without the data included, it is useful to find another means. But it does help fill in missing data in a way that allows for analysis, which should help you complete your research when you need those data points filled in.