I realize that this post will have a limited sphere of interest, but I felt the need to record my reactions to this book for my own sake (so I have a record of it for my sabbatical review) and let others read it if they want to.
I recently finished reading Andrew Hartley’s Christian and Humanist Foundations for Statistical Inference: Religious Control of Statistical Paradigms. I did not agree with the primary conclusions of the book. There are several reasons for this, which I’ll list and elaborate on briefly.
- Hartley overly simplifies the breadth of statistical perspective. He considers two frequentist views and two Bayesian views. There are far more nuances than this among statisticians, which Hartley only briefly acknowledges before trivializing these differences away.
- Hartley views two different forms of Humanism, but only one Christian philosophy. Worse, the view he chooses is clearly not broad enough to include all statisticians. Hartley acknowledges that his Christian philosophy, the so-called “Philosophy of the Law Idea” was developed by two Dutch Calvinist scholars, and seems biased against the many whose philosophy reflects the Arminian/free will view of Christianity. This biased view of Christianity limits the ability to view his conclusions as decisive. In fact, the only mention of freedom of choice or freewill is in connection to humanism.
- Hartley’s view of humanism is defended by many quotes, but most of these come from secondary sources. It is interesting to note what others have said about humanism, but not in place of direct quotes to convince me that these secondary opinions accurately reflect the views of humanists themselves.
- Hartley’s choice for the only Christian perspective is what he terms “Subjective Bayesianism”. Many Christian statisticians would be extremely concerned about this view. Since the Bible is clear that human beings are fallen and imperfect, why should a view that relies on subjective perspectives be endorsed as most Christian? It seems to me this view is possibly most susceptible to abuse by humanists.
- On the other extreme, Hartley’s Direct Frequentism is unfairly maligned on two fronts. First, Hartley makes the claim that Direct Frequentists often confuse the probability of the data given and assumed hypothesis with the probability of a hypothesis given the data. I have read this claim several times, and it is beginning to drive me crazy. This is a first semester stat course mistake that I have never heard a trained statistician make. I warn my students about this all of the time, as well as explaining how to correctly understand what the P-value really is telling them. This critique holds no water with me.
- Second, Hartley assumes that Direct Frequentism is only consistent with mathematically slanted versions of humanism. He never tells me why I, as a Christian who worships a God that created an orderly world, cannot trust the mathematics behind the conclusions of Direct Frequentism. Just because some humanists like it, does that mean I can’t? Hartley never addresses this. I agree with him that ultra-rationalist humanists should support Direct Frequentism or Objective Bayesianism, but those both seem consistent with Christians who trust the orderliness of mathematics (is there anyone who posits that math is fallible and fallen?) over the fallen nature of humans as well.
In all, I find that what he has to say is reasonable, for the most part. His logical tracing of the effects of religious belief are helpful in understanding perspectives on these particular views of statistics, but they are limited. The limitations of not further addressing nuances within statistics, as well as nuances withing Christianity, limit the usefulness of this book. I am not a philosopher, and would not be surprised if Dr. Hartley is a better statistician than I am, as well as a better philosopher. However, this does not excuse the limitations of his work, at least in my mind. I would love to see an expanded work that addresses these issues. The issues that Hartley begins to address here are very important for Christian statisticians to consider, and I appreciate his contribution to the larger discussion.