How to Account for Missing Values in Survey Data

Posted by Anne Schafer on Mar 20, 2019 1:22:43 PM

In any survey, there are situations where a respondent might not answer every question. Whether it’s due to skip logic, or the user is faced with a “select all that apply” question, responses might not always be required. It’s important to take the right approach to in account for missing values, since it may impact the end results of your analyses.

We’ll use a multiple-response variable, or MRV, to illustrate this point, since this is a common question type that doesn’t require a respondent to choose all possible answers. How do you code the MRV for optimal ease-of-use in MarketSight?

Let’s say we have a simple multiple response question like this:

Which kinds of sandwiches do you like? Please pick all that apply:

  • Peanut butter and jelly
  • BLT
  • Reuben
  • Grilled cheese


Your data would look like this:                                                                                                              Your options would be coded like this: 



In the first scenario, we are leaving missing responses blank. It’s important to note that a blank is equivalent to nothing, meaning it is different than a value of zero. As a result, these blanks are not counted towards the total sample size. When you leave the missing responses blank, you’ll ultimately end up with a sample size that is equal to the number of respondents who answered that particular question, instead of the true total respondent count. Your crosstab calculations will reflect this: 



Alternatively, you could take the following approach:

Your data would look like this:                                                                                                        Your options would be coded like this:




In the second scenario, we’re representing those missing responses with a numeric value. This value can be anything—2, 4, 99, etc.—as long as it’s numeric. The end result? You can use the total respondent count as your sample size, and the calculations adjust accordingly.




Both approaches will, of course, work in the MarketSight platform, but we recommend using the second approach. This will ensure that you have a clean variable to run calculations and stat tests with, and will ultimately provide you and your team with more accurate insights.

Topics: survey data, data collection, insights