Most convenient way to detect outliers!?

[Veuve]

What is the most convenient way to detect outliers, beside just looking at the time series graph and find them manually?

[trubador]

The most convenient way to detect outliers is to use a formal "outlier detection algorithm" and many of them can easily be implemented via EViews' programming language. Other than that, visual inspection is not an inferior method to detect outliers, if you use the right tool. Boxplots, for instance, are very useful in identifying "near" or "far" outliers and are also available in EViews' graph types. You can also try EViews' distribution fitting tools to find the most appropriate empirical distribution that fits your data. Then, the outliers would be the ones that fall outside a certain confidence intverval, which you should define depending on the nature of your study. For example, if your data can assumed to be normally distributed, then observations that fall outside the ±3 standard deviation confidence interval "can" be considered as outliers. Please keep in mind that there are numerous efficient outlier detection methods to consider and I'd like to encourage you to do further research.