Data 101

Think of a blank space, devoid of any detail, and in this space lies a singular data point. The question to the reader is whether a singular data point has any relevance? Think long and hard about this proposition prior to reading any further. Think of the value that a single point of data has in the world today, if anything.

To many things and people in this world the single, solitary piece of information is next to useless. Isolated and impossible to quantify or provide qualitative assessment. What if, however, this singular piece of data was a set of coordinates where reaching them meant life or death? Or, what if that single number was a test result value where that number indicates whether or not you have terminal cancer or not? Changes the value and significance drastically doesn’t it?

The value of data is the relevance of what it is related to. Getting the full relationship and position of data is much, much harder than simply collecting it. Take for example, taking the average WOB (Weight On Bit) over a drilled well. If you ran the drill rig through the various formations at this value you would most likely not really drill anywhere. The relationship of the average WOB to efficiently drilling the well isn’t there and is misleading. Take averages compared to downhole WOB over periods of time related to specific formations and you end up with much more useful data. What is the tradeoff? Much more complexity in digging out the truth buried within it.

I was taught by one of my professors an old adage: “asking for a statistician after the data is collected is like asking for a doctor after the patient is dead”. I say this because over the years I have seen good studies of data as well as bad and then the occasional horrible train wrecks pretending to be good science. We all were taught causation is not correlation. We were, somewhere in our math and general sciences. Somewhere the world has forgotten this when it comes to routing out this and that from data sets because I routinely find these mistakes. Sometimes, what should normally be correlation, is masked by other influences.

So, in the end, what I am I really saying. Science and math needs a good philosophical understanding of things. Good philosophy in engineering leads to excellent design and good philosophy in data exploration leads to better comprehension of what the data is really trying to say so quietly in the back ground. This operating philosophy is 'life and death' (data wise) in engineering forensics where gaps exist. There are things that just were not collected and then must be inferred. Anyway, a different topic altogether.

I consider this 101, the very basic first lesson in data. Understand the relevance and how the information relates to the world around it before trying to torture it to death…