Final answer:
The most suitable time-series forecasting method for tracking recent movement in COVID mortality data is likely exponential smoothing with alpha = 0.9, due to its high responsiveness to recent observations. However, this method might also overreact to random fluctuations, so its use should be monitored closely.
Step-by-step explanation:
The selection of a time-series forecasting method that is least biased to track the recent movement of data, and to forecast COVID mortality, depends on the nature of the data being analyzed. Each method has its strengths and weaknesses in different scenarios.
Simple average is not typically responsive to recent changes, as it considers all data points equally. Hence, it might not be the best for tracking fast-changing phenomena like COVID mortality rates.
Exponential smoothing with a high alpha value (such as 0.9) places a lot of emphasis on the most recent observations, making it very responsive to recent changes in the data. This can be both an advantage and a disadvantage; while it captures the most recent trends, it might also respond too strongly to random fluctuations, potentially leading to a biased forecast.
Naive forecasting, which assumes that the latest observation will be the same in the next period, can be suitable when the data does not change much between periods but can be biased if the data has a trend or seasonal component that the naive method ignores.
Linear regression is a more complex model that could consider trends in the data over time, making it potentially less biased if the trend is consistent. However, it might not respond quickly to sudden changes in trends.
Moving average with k=3 smooths out short-term fluctuations by averaging the last three points. It might provide a good balance between responsiveness and smoothing, but it does not respond as quickly as exponential smoothing with a high alpha value to the most recent data movements.
In conclusion, while each method has its merits, the exponential smoothing with alpha = 0.9 could be selected as least biased in tracking the recent movement of data, as it is highly responsive to the latest data. However, the potential for overreaction to random fluctuations should be considered, and thus its application should be carefully monitored.