I am new to Time Series Data and this question is confusing me, as I have received different advice and was wondering if I could request clarification.

I am attempting to test whether the creation of a particular type of school affects the number of rich or poor children in surrounding schools

I have 6 years of data on this, and I was doing a regression along the lines of the number of disadvantaged students, against if there is one of these schools nearby, with a number of other control variables.

Because it was a times series data I was recommended to use a lag of the dependent variable [L.] (since presumably the current number of students will impact next years')

But I was also told to use the D. operator in Stata, by others, to account for any changes in number of such groups between the years, & my search on the web / textbooks has not been overly helpful about which to use in this situation if any at all.


D the difference operator is a statistical shorthand for a case where the contemporaneous effect is perfectly counter-balanced by a lag (delay) effect. Neither a lag or a differencing should be assumed necessary or useful. Analysis using cross-correlation procedures on suitable stationary series can often be useful in identifying an appropriate model.This http://empslocal.ex.ac.uk/people/staff/dbs202/cat/stats/corr.html might be useful in helping you understand the pitfalls of ordinary cross-correlation statistics on the original series. Also http://www.autobox.com/cms/index.php/afs-university/intro-to-forecasting/doc_download/24-regression-vs-box-jenkins discusses the fatal flaw of equally weighting all observations i.e.historical periods in forming a model.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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