What’s Missing at Random (MAR)?

Missing at Random (MAR) is a statistical term indicating that the likelihood of data being missing is related to some of the observed data but not to the missing data itself. This means that the missingness can be explained by the available information. Here are some examples of situations that can be considered MAR:

1. Income and Survey Response: In a household survey, higher-income respondents might be less likely to disclose their exact income. If other variables such as education level, occupation, or home ownership are observed, they can help explain the missing income data. The probability of missing income data is related to these observed variables but not directly to the income itself.
2. Employee Satisfaction Survey: In a company survey on job satisfaction, employees with lower job satisfaction might be less likely to complete the survey. If job satisfaction scores from initial questions or departmental information are available, they can explain the missing data in later questions. The probability of missing responses is related to these initial observations.
3. School Performance Data: In a dataset of student grades, students with lower attendance rates might be more likely to have missing test scores. If attendance data and demographic information are available, they can explain the missing test scores. The probability of missing test scores is related to the observed attendance rates and demographics.
4. Clinical Trial with Side Effects: In a clinical trial, participants experiencing mild side effects might be more likely to drop out and have missing follow-up data. If initial side effect reports and baseline health metrics are recorded, they can explain the missing follow-up information. The probability of missing follow-up data is related to the observed side effects and health metrics.