(a) Which type of smoothing technique is generally better?
In different situations the different smoothing techniques are better. If there exists a strong trend and an up and down seasonality the moving average is the best smoothing technique. But in most of the cases the exponential smoothing with appropriate smoothing parameter works out in all the cases.
(b) How do we determine which of the two smoothing techniques is better?
To determine which smoothing technique is better we have to measure them by using the valid measurements like Mean Absolute Deviation, Mean square error, Mean absolute percentage error, Tracking signal etc. The values of these measurements will inform us among several smoothing technique which one is better.
(c) How can we forecast the value of a time series that contains a secular trend as well as strong seasonal and random variations?
In such a case 1st of all we need to separate out the seasonality. Since random variation can’t be explained, so we need to use some smoothing technique which will give us the trend of the time series. Next using that trend we can calculate the seasonality of each season and thus our analysis will be complete.