The first part is more of an aside. In a variety of contexts, whether for testing Large Deviations Principles or calculating expectations by integrating over the tail, it is useful to know good approximations to the tail of various distributions. In particular, the exact form of the tail of a standard normal distribution is not particularly tractable. The following upper bound is therefore often extremely useful, especially because it is fairly tight, as we will see.
Let 
It is in fact relatively straightforward:
Just by comparing derivatives, we can also show that this bound is fairly tight. In particular:
—-
Now for the second part about CLT. The following question is why I started thinking about various interpretations of CLT in the previous post. Suppose we are trying to prove the Strong Law of Large Numbers for a random variable with 0 mean and unit variance. Suppose we try to use an argument via Borel-Cantelli:
Now we can use our favourite estimate on the tail of a normal distribution.
By Borel-Cantelli, we conclude that with probability 1, eventually 
The question is: was that application of CLT valid? It certainly looks ok, but I claim not. The main problem is that the deviations under discussion fall outside the remit of discussion. CLT gives a limiting expression for deviations on the 
Let’s explain this another way. Let’s take 
But we don’t know how large n has to be before this approximation is vaguely accurate. If in fact it only becomes accurate for 
This looks like an artificial example, but the key is that this problem becomes worse as n grows, (or as we increase the number which currently reads as 1000), and certainly is invalid in the limit. [I find the original explanation about scale of deviation treated by CLT more manageable, but hopefully this further clarifies.]
One solution might be to find some sort of uniform convergence criterion for CLT, ie a (hopefully rapidly decreasing) function 
This is possible, as given by the Berry-Esseen theorem, but even the most careful refinements in the special case where the third moment is bounded fail to give better bounds than
Adding this error term will certainly destroy any hope we had of the sum being finite. Of course, part of the problem is that the supremum in the above definition is certainly not going to be attained at any point under discussion in these post-
By this stage, however, I hope it is clear what the cautionary note is, even if the argument could potentially be patched. CLT is a theorem about standard deviations. Separate principles are required to deal with the case of large deviations. This feels like a strangely ominous note on which to end, but I don’t think there’s much more to say. Do comment below if you think there’s a quick fix to the argument for SLLN presented above.
Related articles
- CLT and Stable Distributions (eventuallyalmosteverywhere.wordpress.com)
- Large Deviations and the CLT (eventuallyalmosteverywhere.wordpress.com)
Eventually Almost Everywhere 
Pingback: Your Questions About Confidence Interval Formula | Find Love Today
The paper to read on this is by Harald Cramer,
Sur un nouveau théorème-limite de la théorie des probabilités,
Actualités Sci. Indust. 736 (1938), 5-23.
Colloque consacre a la theorie des probabilities, vol. 3, Hermann, Paris.
but it’s in French.
Thanks for that, had never chased the reference properly before. Cramer’s theorem is of course the only serious way to address this sort of problem. In this country at least though, the majority of mathematics masters students won’t meet it, whereas everyone sees CLT at A-level, so I guess I was mainly thinking about whether there was a reliable way to refine CLT to suit the purposes of the SLLN via Borel-Cantelli problem. I am increasingly convinced that there isn’t…!
Pingback: The Levy-Khintchine Formula | Eventually Almost Everywhere