Oops, we missed a spot: Comparing data substitution methods for non-random missing survey data in a longitudinal...

Area of Science:

• Psychology
• Statistics
• Data Science

Background:

• Missing data imputation is crucial in longitudinal studies.
• Standard imputation methods may not apply when data are entirely missing.
• Evaluating alternative substitution methods is necessary for reliable data analysis.

Purpose of the Study:

• To compare the accuracy of three missing data imputation methods: mean substitution, Last Observation Carried Forward (LOCF), and regression-predicted values.
• To determine the most reliable method for estimating missing child anxiety data in a longitudinal study.
• To provide guidance for researchers facing missing data challenges.

Main Methods:

• Utilized data from the Ontario COVID-19 and Kids' Mental Health Study, involving 384 parents reporting on child anxiety.
• Applied mean substitution, LOCF, and regression imputation to estimate missing survey items at one and four months post-baseline.
• Conducted Within-Subjects ANOVA and post-hoc analyses to compare imputation methods against actual data.

Main Results:

• Mean substitution significantly differed from actual data.
• Regression-predicted values overestimated the median compared to actual values.
• LOCF produced comparable means and identical medians to actual data, indicating higher accuracy and reliability.

Conclusions:

• LOCF is a more accurate and reliable imputation method than mean substitution or regression-predicted values when advanced methods are not feasible.
• Systematic comparison of alternative imputation methods is vital for informed data processing decisions.
• Findings support the use of LOCF for estimating missing child anxiety data in similar research contexts.