Hi,
This is the first time i am posting, as such, please correct or question me if what i am asking is missing important aspects or is unclear in any way.
Furthermore, a clearcut answer to all or any question is not expected, however, insights and thoughts on the problem/problems at hand are very much appreciated!
Description of data and more general motivation behind questions can be found at the bottom
The questions at hand are the following:
1: I am getting familiar with the user-written pvar program-package of Love and Abrigo (2015), while it is easy to use i am having trouble figuring out/understanding the underlining assumptions.
Does the pvar assume heterogenity or homegenity across units and what about across time? Furthermore, the default option is to use "first difference to remove panel-specific fixed effects", would it be to correct to interpret the estimation method as a model that assumes heterogenity, but when removing these fixed effects one is left with a sort of average effect of how the lagged variable/s impact the dependent variable across units?
2: I haven't managed to find any information regarding a balanced/unbalanced dataset and the impact it as related to panel VAR's. Could anyone shed some light on this?
3: A bit beyond only Stata relevant questions:
While "Normal VAR" needs a couple of checks such as stationarity of the series, i have not come across the same pre-test related to panel VAR's when readying through covered literature. Although i do not understand everything of the math behind the mathematics which deemes panel VAR as an appropiate method, i have the impression that a large number of groups/units even if T if relative short somewhats loosens the assumptions towards non-stationarity, is the correct ?
4: What are the potential pitfalls and important aspects regarding panel VAR's. I have not been able to find a lot of information related to this topic so any links to good sources of information or insights are highly appreciated!
Motivation:
I am currently trying to model the relationship between customers and suppliers in the setting of stock returns.The idea is to apply a Vector Autoregression (VAR) model on individual
customer-supplier relations in order to model the effect how customers stock returns effect supplier returns and vice verca, while also including own lagged returns.
Data description:
The sample consists of around 10,000 pairs of distinct customer-supplier relations, together with monthly stock returns varying in the range of 12 to 260 months of the individual suppliers and customers.
This is the first time i am posting, as such, please correct or question me if what i am asking is missing important aspects or is unclear in any way.
Furthermore, a clearcut answer to all or any question is not expected, however, insights and thoughts on the problem/problems at hand are very much appreciated!
Description of data and more general motivation behind questions can be found at the bottom
The questions at hand are the following:
1: I am getting familiar with the user-written pvar program-package of Love and Abrigo (2015), while it is easy to use i am having trouble figuring out/understanding the underlining assumptions.
Does the pvar assume heterogenity or homegenity across units and what about across time? Furthermore, the default option is to use "first difference to remove panel-specific fixed effects", would it be to correct to interpret the estimation method as a model that assumes heterogenity, but when removing these fixed effects one is left with a sort of average effect of how the lagged variable/s impact the dependent variable across units?
2: I haven't managed to find any information regarding a balanced/unbalanced dataset and the impact it as related to panel VAR's. Could anyone shed some light on this?
3: A bit beyond only Stata relevant questions:
While "Normal VAR" needs a couple of checks such as stationarity of the series, i have not come across the same pre-test related to panel VAR's when readying through covered literature. Although i do not understand everything of the math behind the mathematics which deemes panel VAR as an appropiate method, i have the impression that a large number of groups/units even if T if relative short somewhats loosens the assumptions towards non-stationarity, is the correct ?
4: What are the potential pitfalls and important aspects regarding panel VAR's. I have not been able to find a lot of information related to this topic so any links to good sources of information or insights are highly appreciated!
Motivation:
I am currently trying to model the relationship between customers and suppliers in the setting of stock returns.The idea is to apply a Vector Autoregression (VAR) model on individual
customer-supplier relations in order to model the effect how customers stock returns effect supplier returns and vice verca, while also including own lagged returns.
Data description:
The sample consists of around 10,000 pairs of distinct customer-supplier relations, together with monthly stock returns varying in the range of 12 to 260 months of the individual suppliers and customers.
