DOE

Forum,
Do I have this right for 2 level, orthogonal designs for 1st order models….

Factors of interest can be assessed for significance using among other techniques:  ANOVA or The Interpretability Factor (where the effect is compared to the SE)…are there other ways?
The unknown parameters of the 1st order model (full, fractional, or PB) are composed – after nonsignificant terms are removed -using a least sq method or by calculating the avg difference in effects (+/-) and dividing it by two..again, other ways for this?
Using two center points during screening efforts provides both a source of pure error for significant testing used in model reduction as well allowing the detection of curvature (in a 1st order model).
THANKS!!!!!!!!!!