For example L25 can be used for 6factors and 5 levels for each factor. If i want to design an OA with 3 factors and 5 levels of each factors can I take standard L25 array by reducing its number of columns.
Do you really want to use an L25? You might take a lokk at the Central Composite Design. It allows you to feed a quadratic model with three factors, each of them varied on five levels and it needs just 15 runs. More info in my script (for free, full text download). The L25 has a pretty messy confounding pattern and does not give you the oportunity to isolate the interaction effects. A clean construction, such as CCD might be better. Anyway, it of course depends on you particular case.
Thanks for the reply. Can we just delete the unwanted columns of a standard L25 array, which may suits my requirement? Do I need to check the other properties related to taguchi OA like orthogonality(balancing of columns) etc?