The validity of correlation analysis between finite element model (FEM) and modal test data is strongly affected by three factors, i.e., quality of excitation and measurement points in modal test, FEM reduction methods, and correlation check techniques. A new criterion based on modified mode participation (MMP) for choosing the best excitation point is presented. Comparison between this new criterion and mode participation (MP) criterion is made by using Case 1 with a simple printed circuit board (PCB). The result indicates that this new criterion produces better results. In Case 2, 35 measurement points are selected to perform modal test and correlation analysis while 9 selected in Case 3. System equivalent reduction expansion process (SEREP), modal assurance criteria (MAC), coordinate modal assurance criteria (CoMAC), pseudo orthogonality check (POC) and coordinate orthogonality check (CORTHOG) are used to show the error introduced by modal test in Cases 2 and 3. Case 2 shows that additional errors which cannot be identified by using CoMAC can be found by using CORTHOG. In both Cases 2 and 3, Guyan reduction, improved reduced system (IRS) method, SEREP and Hybrid reduction are compared for accuracy and robustness. The results suggest that the quality of the reduction process is problem dependent. However, the IRS method is an improvement over the Guyan reduction, and the Hybrid reduction is an improvement over the SEREP reduction.