Hello everyone. i have a 25 days data of 47 sites and i am interested in studying the trend of the correlations of 9 variables. My question is; is it possible to do it for a short period such as 25 days? while i tried the rolling regression in EVIEWS where i choose the window size of 10 and step size of 2 i got a notification that there might be collinearity.
Any suggestions? any books ?
Thank you
What do you mean by "rolling regression"? Something like "moving average"?
Why wouldn't it be possible for 25 days? It depends what you want to conclude from that.
It is possible to study trends in correlations of variables over a short period of time, such as 25 days. However, the sample size of 25 days of data for 47 sites may be relatively small, which could affect the power of your analysis and the reliability of your results. Additionally, a short time period may not capture the full range of variation in the data, which could limit the ability to identify meaningful trends.
Regarding your question about collinearity, when you use rolling regression, collinearity can occur when your independent variables are highly correlated. This can lead to unstable and unreliable estimates of the regression coefficients, which can affect the validity of your results.
One possible solution to address collinearity is to use Principal Component Analysis (PCA) to reduce the dimensionality of your independent variables and create new variables that are uncorrelated. This can help to stabilize your regression estimates and improve the reliability of your results.
Another option to consider is to use a different technique for analyzing trends in correlations, such as partial correlation or partial least squares regression. These techniques can also help to control for the effects of other variables and reduce the impact of collinearity.
It's also important to consider the lag time between the 9 variables, as it might affect the correlation results.
Ultimately, I would recommend consulting with a statistician or a researcher with experience in analyzing time-series data to ensure that your analysis is conducted correctly and that your results are interpretable.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling