I would like to better understand the limitations of the MLR in regards to multiperiod data.
Lets say I am looking at financial data for 10 companies, and each of those 10 companies has 5 periods worth of data, can I treat all 100 datapoints the same, or do I need to break it down into individual years?
Additionally, I am noticing that removing a factor will cause another factors p-value to sporadically rise by a large number. My impression of the way a MLR works is that each factor should be independent of the each other, and while the coefficient is supposed to be divided by p number of factors, its relating stats like p-value and t-stat should remain constant if the data for that factor doesn't change
Thanks for the clarity!