When a group Boston College students began a project analyzing data received from UFO sightings, they wanted to receive some insights into the factors that influence alleged UFO sightings, such as weather, movie releases etc. In 2014, the Economist conducted similar research and found that most UFO sightings occur during the so-called drinking hours , between 5-11pm, when people are finishing up their fourth or even fifth bottle of beer. The Economist called this possible connection Encounters of the blurred kind.
All joking aside, Boston College students learned a thing or two about sampling bias. According to the National UFO Reporting Center, UFO sightings have increased exponentially since the organization started keeping track of the reports in 1974. When this organization first opened, people had to pick up the phone and call-in to file a report. Once the internet became mainstream, people could now file reports online by simply filling out a form. This is when the number of reported UFO sightings began to rise. This cheap and convenient reporting method provided more data about sightings, but this increase in data availability fundamentally changed the data set, and any change in data affects the conclusions we can make from this data.
Applying the Lessons Learned From UFO Sightings
If we look beyond the paranormal world, lower costs of data collection are valuable in many ways: We have a lot more data to work with and choose from than ever before. However, we do our due diligence and understand how the data was originally generated and how this influences its value. The sources of data bias can be very discreet and more people within organizations are tasked with interpreting collected data for possible biases.
There are four things data researchers should be on the lookout for:
1. Understanding the History Behind the Data
A lot of times, new data can substantially differ from older data. Businesses must be cautious in terms of how they interpret low-cost, yet rich data obtained from places such as online forums and social media can differ substantially from data obtained from prior sources such as written and telephone surveys. Even though social media is a wonderful source of detailed data about consumer activity and provides unprecedented amounts of information about individuals, not every consumer uses social media and not everybody is honest about what they post online. People intentionally shape their online image into their vision of perfection and could respond differently depending on the medium or visibility of their response. Read also: What’s Wrong With Big Data.
In order to make the lives of data analysts and researchers easier, they must be able to see the data lineage, i.e. the data origin, the systems involved in its collection and the steps that were made during the collection process.
2. More Data Doesn’t Automatically Mean New Data
Even though the amounts of data have increased, the larger problem of sampling errors still persists. The sheer volume of data can provide false comfort: you may think that you have better data when they simply have more weighting of the previous data.
This is where the big in Big Data can fail us. We believe that when a dataset gets to a certain point, it is too big to fail. Furthermore, ingesting and processing new data may require significant processing to transform unstructured data into structured data. It is possible that the new data may require changing business processes to incorporate real-time feeds. These tasks require significant amounts of time, resources and effort.
Before beginning new projects t acquire new data, be sure to assess what information this new data will bring. You must know what new insights you are looking for and, if you don’t know, how can you find out without large investments.
3. Old Data Sources Were Imperfect Too
While you may be tempted to benchmark old data sources with new ones, keep in mind that the old data sources contain flaws as well. So how do you know which data source is better the new one or the old one? There really is no need to choose. Each data source will have its flaws and recognizing them requires experience, but using both can provide more insights than using only one, since one can illuminate the problems of the other.
Be sure to understand how the existing data is limited and how and whether or not a new data source can get around this limitation.
4. Intuition is Still Important
With the increased amount of data from an even greater variety of sources, combining information with intuition is important. Human expertise is still critically important, despite the rise of artificial intelligence and machine learning, in order to see the big picture and understand which part of the picture is reflected in the data.
As data becomes more ambivalent, expertise in data application to solve a specific business problem becomes a key asset.
Undoubtedly, given the vast expanse that is our universe, alien life exists in some way, shape or form. Perhaps even some of the reported UFO sightings may have been actual UFOs. But, let’s be responsible with the data we use because when the aliens do in fact arrive, we need to be ready to warmly receive them. And how will we know when they arrive? Thanks to properly imputed and enriched datasets of course.
