0
$\begingroup$

I have three kind of time series data.

  1. Traffic flow data (jam factor, average speed)
  2. Weather data (for example temperature, humidity, pressure, wind speed, wind direction)
  3. Pollution data (PM 2.5, PM 10)

Data will certainly contain nonlinear characteristics. I mean, for example, in the case of traffic data, there will be a daily cycle (the biggest traffic during the day is 7.00-18.00). There will also be seasonality of data (traffic in the holiday will be smaller)

I would like to extract knowledge from these time series. For example, I would like to know whether increased traffic causes increased pollution. I would like to extract as many of these relations/assosiations as possible.

I'm interested in what kind of mathematical tools can be used to extract needed knowledge using accessible data. I don't have a long experience in maths. I made some research and I have found some tools, but I'm not sure they are suitable for my problem.

  1. Distributed lags models
  2. Granger causality
  3. Threshold autoregressive models

The main questions are. Are these three tools suitable for my problem, and my data? What other tools can be helpful during solving my problem?

$\endgroup$

1 Answer 1

0
$\begingroup$

You can analyze these data using the DSEM functionality of MPlus 8: http://statmodel.com/TimeSeries.shtml

The website has papers and instructional videos on how to apply this technique!

$\endgroup$
3
  • $\begingroup$ I looked at the tool but I couldn't find where 1) it detects and remedies pulses,level shifts,seasonal pulses and or time trends) ; 2) tests for and remedies non-constant model error variance ; 3) tests for and remedies time-varying parameters ; 4) it delivers a family of forecasts for each forecast period using monte carlo/resampling methods. Can it perform these kinds of important analyses ? $\endgroup$
    – IrishStat
    Commented Aug 20, 2017 at 15:03
  • $\begingroup$ 1) you will have to include these effects, just add variables to the long - format datafile to do so; 2) you can specify random error variance at the within-level, 3) same, 4) I don't know what you mean by a family of forecasts, are you talking about extrapolating from the model? Mplus has monte Carlo functionality and can produce a model syntax based on the final values of the model parameters $\endgroup$ Commented Aug 20, 2017 at 16:17
  • $\begingroup$ ok .. I got it the user has to pre-specify the additional effects as contrasted to the software finding them . $\endgroup$
    – IrishStat
    Commented Aug 20, 2017 at 16:56

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.