1
$\begingroup$

In order to analyze the data in presence of seasonality, I used two methods: Proportional hazard model (Cox model) and time series method (Triple Exponential Smoothing (Holt Winters Method)). Now , my question is that what is the difference between them? How should I know which of them is better for my application?

$\endgroup$

1 Answer 1

1
$\begingroup$

Probably neither. You should use your actual data to identify an appropriate model. If you just have a single series you might investigate ARIMA modelling and Intervention Detection procedures. If you have causal series then investigate Transfer Functions/Dynamic Regression and Intervention Detection procedures.

$\endgroup$
4
  • $\begingroup$ Thank you for your answer. I used Holt Winters Method because I saw in some references that if we have "seasonality" AND "trend" in our data we cannot use moving average or single smoothing methods...May I know why do you think those two methods are are not useful for my application? $\endgroup$
    – Fera
    Commented Jul 8, 2015 at 0:09
  • $\begingroup$ I am not familiar with promotional hazard BUT I am very familiar with the shortcomings of assuming any particular structure for a model as parameter estimates are based upon a particular model. Let the data suggest the model.. To prove this point take your data and estimate the hw model and the errors from the model . Present the normal summary statistics e.g. variance /r-sq /aic and your forecasts. I will take your data and submit it to a commercial package AUTOBOX and present the results... Then you decide which approach is best for the data. $\endgroup$
    – IrishStat
    Commented Jul 8, 2015 at 1:21
  • $\begingroup$ Thank you for your suggestion. Regarding ARIMA model, since I am not familiar with that, I have problem with determining p,d and q values for that. Could you briefly help me with that? $\endgroup$
    – Fera
    Commented Jul 13, 2015 at 22:44
  • $\begingroup$ you might want to look at autobox.com/cms/index.php/blog .There is a discussion and some flow charts that might help. $\endgroup$
    – IrishStat
    Commented Jul 14, 2015 at 1:45

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.