Statistical significance is the main criterion for both coefficients and increments to explained variance. If you are working in a new area where there is little known about your topic then small amounts of explained variance may be adequate.

Variance by definition is nonnegative. That means that zero is theoretically its minimum value which indicates that all the data are equal.

Hi Nehra,

I thing this also depends on the problem which you would like to address.

Hello Nagrenda, "How much" variance is noteworthy depends on the variables under consideration. In some cases, explaining 80% of variance might be considered "low," whereas in others, 10% might be considered "high." In other words, context matters.

Statistical significance is one element to consider, but given sufficient sample size, even miniscule values of R-squared (difference) will be statistically different from zero. If one does not have prior experience with the specific variables under consideration, then Jacob Cohen, in his text, Statistical power analysis for the behavioral sciences, made these suggestions as guidelines for MLR:

"Small" effect size: R-squared ~ .02 (e.g., 2%)

"Medium" effect size: R2 ~ .13

"Large" ES: R2 ~ .26

Good luck with your work!