I dont understand why the less assets you have in the porfolio the less the transaction fees. If you have to trade $1k on 2 assets or 100 assets you still trade $1k with lets say .02 fee right?
For example let's say you run some form of Johansen like test and find bunch of assets that are cointegrated. What usually happens is that the weights that you get won't be sparse, it will assign some weight to each of the assets, some will be large and some will be very small. But you can quickly realize that removing the weights that are small does not ruin the cointegration. In other words the standard VECM models will include unnecessary assets with small weights that are not doing anything. That is why sparsity is important, since in a way it let's you find the actual essence of the cointegration, those assets that actually matter.
Godsend
I dont understand why the less assets you have in the porfolio the less the transaction fees. If you have to trade $1k on 2 assets or 100 assets you still trade $1k with lets say .02 fee right?
There may be flat fees associated with each buy and sell order
@@bg271 this is usually not the case on most exchanges
For example let's say you run some form of Johansen like test and find bunch of assets that are cointegrated. What usually happens is that the weights that you get won't be sparse, it will assign some weight to each of the assets, some will be large and some will be very small. But you can quickly realize that removing the weights that are small does not ruin the cointegration. In other words the standard VECM models will include unnecessary assets with small weights that are not doing anything. That is why sparsity is important, since in a way it let's you find the actual essence of the cointegration, those assets that actually matter.