1
$\begingroup$

I am still a rookie at statistics. I am writing a report on crimes and how they affect economic prosperity. I have yearly data for 30 years from 1 country, where number of crimes in a year is my independent variable and GDP growrth is my dependent variable. I have 3 other control variables as external invtment, total trade and infrastructure.

Now my key question is can I use multivariate time series? if not, which is the best regression in this case?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Multivariate time series (ARMAX models) are a generalization of "regression models" as you know them. The ARMAX model is also sometimes called a Transfer Function model. A final model can include appropriate contemporaneous and lag structures for your pre-specified causal series and potentially both an ARIMA structure to reflect omitted stochastic series and empirically identified Pulse/Level Shift/Time Trends/Seasonal Pulses which can reflect/proxy omitted deterministic effects/series. Some data sets require either power transforms or Generalized least Squares to render an error process invariant over time.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you very much Sir. Do I need to first check whether my data is stationery or non-stationery and the run the regressions? $\endgroup$
    – MIS87
    Commented Nov 11, 2016 at 14:45
  • $\begingroup$ Pleas use onlinecourses.science.psu.edu/stat510/node/75 as a possible way to identify the model which will include any needed non-stationarity . There are cases where both x and y are themselves non-stationary BUT the Transfer Model does not include any differencing operators. $\endgroup$
    – IrishStat
    Commented Nov 13, 2016 at 0:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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