Question 2.

Thanks to Alfred Marshall, economists typically graph supply and demand with Q on the horizontal axis and P on the vertical axis, although we usually model supply and demand quantities as functions of price rather than vice versa.  So to graph this problem, you would invert the supply and demand equations to get willingness-to-pay (WTP) and willingness-to-sell (WTS) equations:

    WTP = P_D = 100 - Q
    WTS = P_S = 25 + 0.5Q
 

To solve for point elasticities, calculate (dQ/dP)(P/Q).  At equilibrium, P=50 and Q=50, so the point elasticities happen to be the demand and supply slopes: E_D = -1; E_S = +2

Consumer and producer surpluses are based on the assumption that the market has no price discrimination: the price of every unit is the same for everyone.

Consumer surplus is the collective dollar amount all consumers would hypothetically be willing to pay above what they do pay (PxQ) for the equilibrium quantity.  Consumers are willing to pay prices ranging from $50 to $100 per unit, but they all pay just $50.  Graphically, this is the cyan triangle, which has area (50 x 50)/2 = $1,250.  (Area of a triengle is one-half base times height, remember?)

Producer surplus is the collective dollar amount all producers receive above the total minimum amount they would hypothetically be willing to accept for the equilibrium quantity.  Sellers are willing to sell at prices ranging from $25 to $50 per unit, but they all get $50.  Graphically, this is the yellow triangle, which has area (50x25)/2 = $625.
 

Question 3:

Okay, you're supposed to figure out the story this equation is trying to tell.  The regression coefficient on years in the league (YRS) is positive but not statistically significant at the 90 percent level.  You generally look for t-values (absolute values) greater than 1.68 (= 90 percent significance) or 1.96 (= 95 percent significance).  A t-value of 1.05 indicates that the coefficient on YRS isn't significantly different from zero.  All the other regression coefficients are significant.

SALARY = 105.4 + 42.6 (YRS) + 6.6 (PTS) + 4.1 (PNLTY) - 0.026 (PNLTYSQ)
           [3.66]  [1.05]      [2.41]      [4.60]        [-2.07]

                               N = 244 players                 R-Square = 68.6

N, the number of observations, is plenty high enough.  If N<30 you should refer to a t-table for appropriate (higher) t-statistics associated with desired significance levels.  R-Square represents the proportion of total variation in salaries explained by this model.

For the final question, note that the model is a quadratic function of penalty minutes, suggesting that there is a hypothesized optimum number of penalty minutes.  If a player never gets penalized, he's probably not aggressive enough to be effective.  On the other hand, he also hurts the team if he spends too much time in the penalty box.  The positive coefficient on PNLTY and the negative coefficient on PNLTYSQ support this hypothesis.  Both are significant at the 95+ percent level.

What is the salary-maximizing number of penalty minutes?  Take the partial derivative of salary with respect to penalty minutes, set is equal to zero, and solve for optimal P*:

     dS/dP = 4.1 - 0.052P* = 0    P* = 4.1/0.052 = 78.85 minutes.