As a PhD statistician, I love a good data-driven story. Increasing, people in politics, business, and academia are looking for decisions that are based on what the data say. Leavitt's research is engaging and accessible and I, for the most part, enjoyed this book.
HOWEVER, without exception, Leavitt presents his findings as gospel and continually fails to acknowledge the limitations of his methods and his data. He mentions his use of linear regression to obtain his results, but fails to mention the limitations of this method (e.g., results are probabilistic, results are based on model assumptions which may be entirely incorrect). His results obtained from this method sometimes also appear to tell too convenient of a story and seem to be cherry-picked. Moreover, all his results are based on single data sets and may not be as universal as he would like. Finally, he often takes one result (e.g., reading to your kids does not affect their standardized test scores) and makes huge, sweeping generalizations that lead you to believe that reading to your kids doesn't have any affect on any outcome of interest and that you're a bad (or naive) parent for even trying.
These are dangerous practices, though I can see why he does what he does - making all sort of caveats would water-down his findings and make his book less sensational. Nevertheless, he runs the risk of misleading his readers. Judging from the comments posted here so far from people who assume these conclusions are certain, I would say he's succeeded in this endeavor.
Report Inappropriate Content