Long-memory analysis of time series with missing values

Area of Science:

• Time Series Analysis
• Statistical Modeling
• Data Science

Background:

• Accurate estimation of long memory is crucial in various fields.
• Missing data presents a significant challenge in time series analysis.
• Existing methods for handling missing data may impact long memory estimation.

Purpose of the Study:

• To investigate the impact of common gap-filling techniques on long memory estimation.
• To compare the performance of interpolation, random, and mean filling methods.
• To identify reliable methods for long memory estimation in the presence of missing data.

Main Methods:

• Numerical simulations were employed to assess the effects of gap-filling.
• Three gap-filling techniques were tested: interpolation, random filling, and mean filling.
• The study analyzed both persistent and antipersistent time series.

Main Results:

• Gap-filling techniques introduce significant deviations in scaling behavior for both persistent and antipersistent time series.
• The effectiveness of gap-filling methods varies depending on the time series characteristics and gap size.
• Interpolation demonstrated potential reliability for persistent time series when gaps are smaller than the scale of interest.

Conclusions:

• Simple gap-filling methods can compromise the accuracy of long memory estimation.
• Careful consideration of the chosen gap-filling technique is necessary for reliable time series analysis.
• Interpolation may be a viable option for persistent time series under specific conditions.