1
$\begingroup$

I usually perform sensitivity analysis on physical systems. So, for one configuration I have 1 answer and I can build for example a design of experiment and compute the sensitivity of each parameters.

But this time, i would like to perform a sensitivity analysis on a time series : if I change a parameter, it can have an impact later, not immediately. I would like to analyse data on a long period and not only in short term. Do you know existing methods that perform such analysis ?

Or do I have to transform my time series into "slots of period" and study each slot separately ?

$\endgroup$

1 Answer 1

1
$\begingroup$

Performing sensitivity analysis on time series data requires a different approach. One way to do this is by using dynamic linear models (DLMs). DLMs can capture the effect of a parameter change over time and can be used to perform sensitivity analysis.

Another way is to use a time series model such as autoregressive integrated moving average (ARIMA) or seasonal autoregressive integrated moving average (SARIMA) to analyze the data. These models can be used to identify the impact of a parameter change on the time series over a longer period.

Otherwise, you could also split your time series into smaller time periods and analyze each period separately. This approach could also help identify the impact of a parameter change over a specific time period. However, it is important to keep in mind that this approach may not capture the full picture of the time series behavior.

$\endgroup$
1
  • $\begingroup$ Thank you, I'm going to compare those manners. I also discovered Lagged Response Analysis : scipy.signal.correlation_lags but I don't know yet if it is a good stuf. In my case, I try to find a relationship between R&D investments and incomes of a company... $\endgroup$
    – lelorrain7
    Sep 4 at 10:56

Your Answer

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

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