Regression analysis and time series data

Question

• I'm trying to model the effect of flow at entry and flow at exit on the produced energy in a thermal power plant ,The data is collected every 2h for two years,in my internship they want me to model these time series data but the supervisor told me to go for a simple linear regression without any lagged values , and then run a time series model, i failed to create a regression model because the data suffered so much from heteroscedasticity and serial correlation , when i used WLS i wasn't able to correct for heteroscedasticity and serial correlation , and when i used NLLS with one lagged features , i was able to get rid of serial correlation but not heteroscedasticity , because the data is so noisy .
• I don't know the goal from this approach ?
• and what is the goal from this regression analysis ,do i need to find a blue linear regression ?

Answer 1

You seem to be confused as to what time series data is. To be clear - using time series data does not necessarily mean that it has a lag effect. Time series data is data that has been collected in a time chronological order and does not exclude data on certain grounds (versus what panel data would do).

Heteroscedasticity could be due to missing variables or extreme outliers in your data. You can handle for both whilst keeping a time series approach. How you wish to handle it is up to you/your company/your professors advice/what data you have at hand.