deseasonalize data

[高佳]

Hi all,

As is shown in the above snapshot, the authors claime that the data of variable "OpInc", which is quarterly , "is deseasonalized", and I don't know how.
any suggestion and advice would be deeply appreciated.

[William Lisowski]

Apparently you are unfamiliar with seasonality in time series data. The Wikipedia article on seasonality is a place to start. Within Stata, the output of search seasonality includes, among other things, a link to a Stata Journal article by Nick Cox (Stata Tip 76: Separating Seasonal Time Series) and a pointer to the variants of the tsfilter command. With regard to the latter, the tsfilter documentation in the Stata Time-Series Reference Manual PDF is much more comprehensive than the terse results in the help files linked to from search.

[Nick Cox]

William gives good advice. But note that there many ways to deseasonalize data, and we can't tell you which was used in that paper (unless freakishly someone has read that paper and recognises the output).

[高佳]

Apparently you are unfamiliar with seasonality in time series data. The Wikipedia article on seasonality is a place to start. Within Stata, the output of search seasonality includes, among other things, a link to a Stata Journal article by Nick Cox (Stata Tip 76: Separating Seasonal Time Series) and a pointer to the variants of the tsfilter command. With regard to the latter, the tsfilter documentation in the Stata Time-Series Reference Manual PDF is much more comprehensive than the terse results in the help files linked to from search.

Thanks William, I'm really a rookie to STATA and econometric. Your advice do help me a lot!!!

This really is a great forum!

[高佳]

William gives good advice. But note that there many ways to deseasonalize data, and we can't tell you which was used in that paper (unless freakishly someone has read that paper and recognises the output).

I agree with you Nick,
have already emailed the author with some other questions like how to deal with the large amount of negative values when make the log transformation before I came here to post, but there still nothing back.

[Nick Cox]

Negative logarithms are not a problem. Negative values rule out logarithms as such.

[高佳]

Negative logarithms are not a problem. Negative values rule out logarithms as such.

but can you still take log even if more than half of your data is negative??? or is there some standard ways to process the data to make it suitable for logarithm? I know you can replace 0 as 0.001 or simply plus 1 to 0, but my data has such big absolute value that seem impossible to deal it this way.

[Nick Cox]

Many threads here on Statalist and on Cross Validated about this. Google "cube root transformation Stata" and then "neglog transformation"