This is a recap of the recommended 'brute-force' fitting procedure to find the baselines B_ew and B_ne, which is equivalent to finding Q_ew and Q_ns (defined in the first few lines of section 8.4 in lab3 writeup). 1. Follow the first few lines of section 8.4.3 to find the best value for Q_ew assuming that Q_ns = 0. That is, generate a series of 'guessed values' for Q_ew. For each one, least-squaes fit for the unknown coefficients ( A, B) in eqn 12 and derive the sum-of-squared residualsm S^2. Plot S^2 vs Q_ew; the least-squares fit for Q_ew is that value for whjich S^2 is minimized. Issues: Cover a large enough range of Q_ew in you guesses so that you find the true minimum, not just a local minimum. Also, make then guessed value clopse together enough so that you define the minimum well. 2. Repeat the above for a 2-d grid of guessed values for Q_ew and Q_ns and find the global minimum for S^2 in this 2d space. Call this minimum value S_min^2. The guessed values for Q_ew and Q_ns at this global minimum are the best-fit values. Call these Q_ew* and Q_ns*. Define \Delta Q_ew = Q_ew - Q_ew* and \Delta Q_ns = Q_ns - Q_ns* 3. Deriving the uncertainties in the best-fit values: 3.1. Convert your 2-d grid of S^2 into a 2-d grid of \Delta S^2, where \Delta S^2 = S^2 - S_min^2. Then the following is true: \Delta S^2 = [\Delta Q]^T ## [\alpha] ## [Delta Q], where [\Delta Q] is tbe column vector of quessed values minus the best-fit values squared, i.e. [\Delta Q] = [ \Delta Q_ew \Delta Q_ns] and [\alpha] is the curvature matrix (equation 2.4a in 'least-squares-lite'), which is 2x2 and symmetric so it has 3 independent parameters--2 diagonal and 1 off-diagonal element. 3.2. Extract approximate values for the elements of \alpha by hand. Use your native intelligence to do this in an easy way. 3.3. Having found [\alpha], find the covariance matrix by taking its inverse. 3.4. Derive the uncertainties in Q_ew and Q_ns using equation 3.7 in 'lsfit-lite'.