> Climbing to a higher vantage point can also unlock new forms of extraordinary that you might have never noticed before. I read an article once about how the amount of work to get into the top tier in a single area is astronomical, but the amount of work to become top tier in a combination of 2-3 fields is attainable by almost anyone. For example, becoming a top tier statistician is hard. But becoming a top tier sta…
I think this is general work advice and in that sense I think it makes sense; it holds even more for {x, programming}.
From my perspective, specificity and personality is an important part of the creative process towards becoming an expert or otherwise accomplished in a field, or at least, a field that has some inventiveness to it. In other words, focusing on even fewer things than a field such as statistics.
To use the statistics example, I view statistics, or at least applied statistics, from the point of view of multivariate statistics. This makes me quite ignorant towards frequentist statistics (no, I am not about to promote Bayesian statistics, which in my opinion is the word that often refers simply to the Bayes Theorem in the context of frequentist statistics) but with very clear reasons:
1. Normality doesn't really come into play in multivariate stats unless we are sampling randomly and without context. In multivariate stats, we actually tend to either use an exhaustive dataset or we sample contextually, or maybe uniformly along a geographical feature such as a riverbed.
2. Dependence of variables is why we resort to multivariate statistics in the first place.
3. Sample size, similar to 1.), is often not applicable. It does apply in the sense that a small matrix gives few insights. However, replicates in an experiment in frequentist statistics are used as a measure of internal variation in relation to between group variation. In multivariate stats you can do the analysis in complete ignorance of whether there are groups at all, or, you can use the groups in a similar way and consider replicates. But there are not hard conventions like having at least three replicates in order to define what variation means in the first place. In multivariate stats variation is a measure of variation between variables, not groups, and the principal components order variation in terms of new, virtual variables via eigenvectors.
4. For a combination of these reasons, the general application of different types of statistics, specifically frequentist statistics as opposed to multivariate statistics tend to change the whole field you are working in to start off with.
Now, the reason for my elaborate story about differences in statistical approaches is that to be an expert in anything, I personally think is a journey towards discovering what makes your interests peculiar, or if you prefer another word, special. (Also, it depends on which experts you happen to be exposed to.) So, yes, of course it should be hard to be one of the best in anything, but I think a reasonable approach is to find the peculiarities of what you could become deeply committed to.
Frequentist statistics is contextually difficult for me to be focused on, and that makes it difficult. However, once the central limit theorem gets introduced, I instead would venture towards it and then once again forget whether I even can ever have enough time to properly study sampling. After all, I wasn't the one to happen to work for Guinness. Beer, now that would be a good reason to study sampling.