I'm using multiple time series measured daily from 2015 to 2021. but records for some days are missing for all time series. How can I impute the records for missing days?
The choice of method depends on the nature of the missing data and the characteristics of the time series. For example, if the missing values are missing at random, simple methods such as mean imputation or LOCF may be sufficient, while if the missing values are not missing at random, more complex methods such as multiple imputation may be needed. It is also important to consider the potential impact of the imputation method on the results of the analysis.
I disagree with Dinesh Kumar. Under the assumptions of missing completely at random (MCAR) or missing at random (MAR) data, full information maximum likelihood (FIML) or multiple imputation may be used. Under missing not at random (MNAR) data, these techniques can lead to bias. Also, mean imputation or LOCF are no longer recommended in the missing data literature (e.g., Enders, 2022). Mean imputation can lower the variance and LOCF can obviously introduce bias in longitudinal data.
Enders, C. K. (2022). Applied missing data analysis. Guilford Publications.
I agree with concerns raised by Christian. However, when you have no options then you can use points mentioned by me. Do you agree on this Christian Geiser?
Dinesh Kumar I'm not sure I understand what you mean by "no options." Even for MNAR data, you do have options that are better than mean imputation and LOCF (which are almost never appropriate). I agree that using multiple imputation or FIML may be better than using other methods (including mean imputation, LOCF, listwise deletion etc.), even when data are MNAR.
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing