My supervisor recommended that I use GMM for my paper and after reading more about the technique, I'm admittedly very confused on whether it works with my data or not?
My assumptions is that I have endogeneity, heteroskedasticity, and cross-sectional dependence.
Many thanks
Using a Gaussian Mixture Model (GMM) when you have N=11N = 11N=11 (features) and T=25T = 25T=25 (samples) can be challenging, but it might be appropriate depending on the context and goals of your analysis. Here are a few considerations:
In summary, while it’s possible to use GMM with N=11N = 11N=11 and T=25T = 25T=25, you should carefully consider the potential for overfitting and ensure you use methods to mitigate this risk. Experiment with different configurations and validate your model to ensure it’s providing meaningful insights.
Safiul Haque Chowdhury Thank you for taking the time to answer my query! I'm afraid I will probably be at risk of overfitting due to the nature of my research (macroeconomic study which incorporates many different control variables) but I think I can make use of PCA to reduce the complexity of some of my sub-control variables.
How can I know if my model is overfitting?
Once again, many thanks for your response.
Nour A. Hello, Thank you for the response. Before you reduce overfitting. First you have to be sure if your model is overfitting or not. To determine Overfitting. You have to generate learning curve. By seeing the learning curve you will know whether the model is overfitting or not.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Social Media