Not exactly innovation research, but probably relevant for any academic discipline: How we are biased in our believe in empirical results through studies that remain unpublished, because they fail to prove significant results and hence do not get published:
Ben Goldacre: What doctors don’t know about the drugs they prescribe
„Even a large estimated coefficient would not be significant if the standard deviation is also very large. Generally, the lowest ratio to be statistically significant would be about 1.65 (though there are other factors that determine this). A high level of significance would require a ratio closer to 2.5. Thus, one can look at any table and expect significant coefficients to be around twice the value of their standard deviation (or more).“
Russel (Russ) Coff (2012), member or former member of editorial boards including Academy of Management Journal, Academy of Management Review, Organization Science, Strategic Management Journal, and Strategic Organization.