Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite

I have the annual time series data of growth variable $X$ for 50 years. Most of the values for the variable $X$ are less than 10%. The exception are two values that are around 30%. How do I treat these two values in the regression?

I would like to seek your suggestion in this regard.

share|improve this question

1 Answer

up vote 1 down vote

This would depend on what you know about the outliers. They should not be removed unless there is some obvious error. They could be indicative of interventions. First step is to understand why you have them. Then you can decide how to treat them. In addition to the covariate you might want to include autoregressive terms in the model. You may also want to investigate how the outliers affect the regression parameter for X (perhaps by using influence functions).

share|improve this answer
Thanks Michael for the prompt response. The data are not due to error, but they are due to the intervention. I also included AR terms in the model.As to the influence functions, these two observation lie above the cutoff point suggested by Cooks distance. Is it okay to remove these observations in time series analysis? – nic Jul 15 '12 at 14:12
@nic: If you know that they are due to an intervention, then add an indicator variable for the intervention. You'd have to decide what kind of impact the intervention would have: is it 1 in the year of the intervention and 0 otherwise, or is it 1 in the year of the intervention, 0.5 in the following year, and 0 otherwise, or ... The bottom line is that you have determined that the data is in fact legitimate and explainable, so add that explanation to your model. – Wayne Aug 14 '12 at 14:04

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.