ケインズの経済学、VaRのシミュレーション

最近英国の経済学者JMケインズの「雇用・利子および貨幣の一般理論」を読んでいます。分かりづらい点ひとつ記録します。
Keynes discusses in chapter 6 the definition of income for a certain time period in terms of output sold(A), user cost(U), and factor cost(F). The user cost is a combination of the goods bought from suppliers(A1) and costs due to the cost of running machinery or the value of goods produced at an earlier time which contributed to A in this period. A-U-F is the income for the entrepreneur. The aggregate income for the community is considered to be A-U. What I don't understand is his equating "the marginal proceeds to the marginal factor-cost at every point on the aggregate supply curve". In other words:
delta(N) = delta(Aw)-delta(Uw)
This means that the change in labor cost would exactly offset the change in income before labor cost for the entrepreneur. Perhaps the point here is that the entrepreneur would increase production (thus N) up to the point where the above holds, as up to that point it is profitable to produce more. I assume that the supply curve is a graph of supply vs. price, but this has not been confirmed.
Another area of confusion was the statement that "the aggregate supply function is linear with a slope given by the reciprocal of the money wage". It makes sense that, the lower the wage, the easier it is for the entrepreneur to increase production with a price increase, but I do not see the mathematics behind this. If Zw is number of wage units in aggregate supply and W the wage unit, The change in Zw with respect to price can be written:
dZw/dp = (1/W)(dZ/dp)
If we know dZ/dp to equal 1, then the statement of Keynes holds in terms of the aggregate supply in terms of wage units.

Another area of confusion is in the "Monte Carlo Simulation" section of the Value at Risk exposition I am reading.
In discussing the bias and mean-squared error(MSE) of the quantiles, the vector norm is used. However, I am not sure why vector quantities are involved for quantiles. What does each coordinate of the vector indicate? The text indicates that the simulation involves 1000 samples of 2000 observations. Based on the divisor of the vector norm, it seems that the vectors consist of 2000 coordinates. This especially makes no sense, as the earlier bias and MSE computations, also for simulations of 1000 samples of 2000 observations, were scalar quantities. Thus, I do not see the point of making them be vectors of 2000 degrees in this second simulation.