thanks, I like your simplicity and clarity of your explanations. I feel you need to explain more on the second and third strategies for dealing with unobserved heterogeneity. Could Fixed effects, demeaning and differencing.... be an option here?
Thanks for this. I have a question. The Regression model for clustered data contains double indexing i.e. each cluster gets its own regression line. Is not that enough to account for heterogeneity in the data. Why do we need to add a separate term to account for unobserved heterogeneity (i.e. aj)?
If you estimate as separate line for each cluster, then you are addressing unobserved heterogeneity. But that is not typical, at least in management where clusters tend to be small. Instead, we estimate the same slope for all groups and allow only he intercept to vary.
@@mronkko Thank you for the answer. What I was able to understand from your answer is that in ideal conditions a separate line for each cluster is enough to address the unobserved heterogeneity (i.e. we can do away with the 'aj' term) In other words: Assuming that the dataset is large enough and every assumption is satisfied by the dataset except the ones related to unobserved heterogeneity: 1. Will a multilevel model without the 'aj' term be enough to account for unobserved heterogeneity?
Hi! Thank you so much for this video. I have quick question - can these methods be used to eliminate unobserved heterogeneity purely on cross sectional data? For example if I have country level data, but I have reason to believe that different cities have unobserved characteristics that affect my y. I have only looked into fixed effects, but I guess that fixed effects modelling is only for panel data and not cross sectional data. Thank you once again
Thank you very much for your videos! Your explanations are so clear and simple!
thanks, I like your simplicity and clarity of your explanations. I feel you need to explain more on the second and third strategies for dealing with unobserved heterogeneity. Could Fixed effects, demeaning and differencing.... be an option here?
Thanks for this.
I have a question. The Regression model for clustered data contains double indexing i.e. each cluster gets its own regression line. Is not that enough to account for heterogeneity in the data. Why do we need to add a separate term to account for unobserved heterogeneity (i.e. aj)?
If you estimate as separate line for each cluster, then you are addressing unobserved heterogeneity. But that is not typical, at least in management where clusters tend to be small. Instead, we estimate the same slope for all groups and allow only he intercept to vary.
@@mronkko Thank you for the answer.
What I was able to understand from your answer is that in ideal conditions a separate line for each cluster is enough to address the unobserved heterogeneity (i.e. we can do away with the 'aj' term)
In other words:
Assuming that the dataset is large enough and every assumption is satisfied by the dataset except the ones related to unobserved heterogeneity:
1. Will a multilevel model without the 'aj' term be enough to account for unobserved heterogeneity?
Hi! Thank you so much for this video. I have quick question - can these methods be used to eliminate unobserved heterogeneity purely on cross sectional data? For example if I have country level data, but I have reason to believe that different cities have unobserved characteristics that affect my y. I have only looked into fixed effects, but I guess that fixed effects modelling is only for panel data and not cross sectional data. Thank you once again
@@mronkko Thank you SO much for the quick and helpful reply! You're the best :)
wow this video was super super super helpful!!!!
I am happy that you found it helpful!