Hello Adil,

The values of unstandardized regression coefficients are heavily dependent upon the scale of that predictor/independent variable (IV) and the dependent variable (DV).

Remember the interpretation: "B" refers to the expected change in Y (the DV) per unit change in the IV, holding all other IVs constant.

A simple example: Data from U.S. Baseball players in 2013 indicates that the correlation of how many home runs a player hit (the IV) correlated with income (the DV) at r = .251.

The unstandardized regression coefficient is:

0.002, if income is scaled in 10,000,000 dollar units (and is stat. significant)

0.220, if income is scaled in 100,000 dollar units (and is stat. significant)

21967.73, if income is scaled in individual dollar units (and is stat. significant).

In each case, the standardized regression coefficient is the same, 0.251.

Good luck with your work!

Adil -

You said "...the very less B value does not make any sense...." Does it give you a strange result for predicted y? Predicted y is what matters.

Cheers - Jim

PS - By the way, a lone p-value is not useful. Sample size matters. Standard deviation matters.

The predictions and estimate variances of the prediction errors are of use to you here.

Does it all make sense from a subject matter perspective?

Well, computer calculations are done in double precision so small numbers are okay up to a point… 1. Multiply DENSITY BY A CONSTAT e.g. 10, 100, 1000 etc & RERUN ANALYSIS AGAIN (P-value is a ratio of two numbers, so it’s size depends on relative sizes of your numerator & denominator 2. Small numbers must be consistent with the possible detection level of your measuring INSTRUMENTS… 3. Data distribution consideration may need association calculation via Rank correlation…Rank correlation does not assume Linear relationship… it is shape independent…