Introduction to missr

What is missing data?

Missing data is defined as “data that we intended to collect, but for one reason or another were unable to” [1].

The problem of missing data

Missing data represents a loss of information, and so reduces the statistical power of a study [2].

Why can I not just impute the missing data?

Different methods for handling missing data (e.g. mean imputation) can introduce bias when estimating statistics [3]. Therefore, to produce unbiased estimates, one must choose the correct method. In order to do this, one must understand and categorise the pattern of missingness in the dataset. Having categorised the missingness as MCAR, MAR, or MNAR (see vignette("background")) one can then choose the correct method for handling the missing data without introducing bias.

How does missr work?

This package provides three functions — mcar(), mar(), and mnar() — to help you categorise the missingness in a dataset.

This package also provides several toy datasets:

• MCAR
  • animalhealth
  • pollutionlevels
• MAR
  • healthcheck
• MNAR
  • companydata
• No missing data
  • testscores

Why was missr created?

There is a lack of information and software on practically categorising missing data.

References

[1] Carpenter JR, Kenward MG. Missing Data in Randomised Controlled Trials: A Practical Guide. Health Technology Assessment Methodology Programme; 2007.

[2] Pham TM, Pandis N, White IR. Missing data: Issues, concepts, methods. Seminars in Orthodontics. 2024;30(1):37-44. Statis- tics every orthodontist should know.

[3] van Buuren S. Flexible Imputation of Missing Data, Second Edition. Chapman and Hall/CRC; 2018.