how to measure Model performance in a uni-variate time series data?

Hi,

i have univariate time series data sets of water levels named as: observed data and simulated data .

i want to transform the simulated data in a way that it represents the variability and mean of the observation.

i.e. need to fix simulated data as may be like 95% variability and the mean of the obsrvation.

Or even better, how can i apply a transfer function that would transform the simulation into the observation?

i have joined them in a timetable. the data is attached as a mat file . first column is timestamp, second is observed data, and third column is simulated data.

please guide me how can i present the simulated /model performance?

what other measures should i consider for this ?

i have calculated RMSE,and nasch sutcliffe efficiency coefficient as well,(files of codes are attached). but its hard to interpret the results.

i need some guidance on this too.

looking forward for your guidance.

[Peter Perkins]

This doesn't seem like a MATLAB question. If you have simulated data, they must have come from a simulation model. It aseems like your questions ought to be

• Is the model producing data "like" the real data?
• If not, how should the simulation model be changed?

Looking at the simulated data for the second question seems like you are too late.

I don't think anyone will be able to help with this question as posed, because only you know the details of your data and your model.

[bushra raza]

thank you for the reply.

i have raised similar question with more clarity i guess on

https://www.mathworks.com/matlabcentral/answers/515695-how-well-the-modeled-data-fit-to-the-observed-data

i am still eager to show 95% variability of simulated data versus observed data.

i dont know about simulation model, but i just have its values.

if simulated data represent the past years records say 1963-2016 well, then it means simulated data can be used for future projections. as simulated data is upto 2099.