Multiple
regression analysis - a common concern from Independent Study (IS) students
Student A:
I have generated my multiple regression report, but it shows that all the
b-values (of the x variables) have almost zero values and that the p-values are
so large that the null hypotheses of b values being zero cannot be rejected. Is
it a big problem? What could I do now?
Professor
B: It needs not be a problem. There are
three actions you could consider:
i. keep
analyzing your existing data with your existing regression formula for your assignment work [remember: no major findings is still a finding];
ii. revise mildly
your multiple regression formula with, e.g., different variables, in an
exploratory mode to see if something more interesting can be discovered;
iii.
conduct some complementary data analysis using scatter diagrams, bar charts, or
pivot tables, etc., to enrich your data analysis.
All the three
actions are OK.
Student A:
I don't want to be OK; I need to achieve high score.
Professor B:
ic; in this case, consider action (ii) or (iii); action (iii) tends to have a
higher ROI on your effort made. You could also consider to adopt both (ii) and
(iii), if you like and have time to do so. If you take action (ii), you need to update your literature review to support your revised multiple regression formula setting.
