Methods of Data Analysis
Broadly speaking, combining the traditional and computational methods of data analysis yield a well-rounded result. In the traditional method, the emotional or human factors are captured whereas the algorithms from the computational method are limited in scope and give a measurement report omitting the emotional or social variables. An analysis based solely on one of these methods leads to skewed results in the direction of the method.
I became interested in researching the approaches that are used in data analysis because of Brooks’ Banking Executive example that appears in his second article, “What Data Can’t Do”. This example has two interesting aspects. First, it expands on Brooks’ view from his first article, “The Philosophy of Data”, where he indicates that statistical analysis reveals new patterns that humans fail to notice. Second, it shows that consideration of both the computational and human aspects leads to an informed decision making process. Hence, Brooks’ Banking Executive example highlights the important questions about interdependence of computational data analysis, human factors and real world decision making.
Brooks’ two articles explore data analysis from various angles. His first article leans towards the computational method and shows that ideals that humans hold dear often negatively impact the analysis of data. This is because these ideals simply may not reflect reality. Brooks proves this in his first article with the analysis from ‘Gilovich, Tversky and Vallone’, “that a player who has made six consecutive foul shots has the same chance of making his seventh as if he had missed the previous six”. Hence, people’s intuition that a player can have hot and cold streaks in a game is incorrect. In Brooks’ second article, in the Bank Executive example, he vaguely mentions that the executive was not oblivious to the data from the computational analysis. Then, Brooks went on to stress the executive’s decision to “remain in the weak economy and ride out any potential crisis, even with short-term costs” and that this decision was based on the emotional and trust connection the Bank had established with the people of Italy. It is this example that illuminates the power of decisions made when both computational and traditional methods of analysis work together.
One of the pluses of computational analysis is the ability to analyze large datasets quickly. Stewart at al acknowledges this in the research when they said, “investigators can now rely upon alternative sources and techniques to corroborate information about public health events”. However, can one rely on this result? Researchers have found that more is needed to garner better results. Stewart acknowledges that although the statistical pattern recognition algorithm was good for their research, it raises more questions about the quality of the variables used to determine the pattern. Hence the team has to extend their research to consult with “domain experts” Stewart (8). Here, a team of researchers tackle a large dataset and are held back because one method of analysis did not give a comprehensive result.
In Lewis et al research, the analysis is questioned when large datasets are computationally analyzed, because software has its limitations. This research, dealing with analysis of data over the internet, shows that “non-traditional variations are needed to cope with the unique nature of the internet and its content” Lewis (35) while “algorithmic analysis of content remain limited in their capacity to understand latent meanings or the subtleties of human language” Lewis (35). Software is not able to interpret every human instinct, sudden emotional outburst or behavior that humans express in slang. Human social culture changes rapidly and makes analyzing data written on the web more complex, since the meaning of a slang word today can have a completely different meaning tomorrow. This makes it more difficult to rely solely on the computational method.
Both Stewart et al and Lewis et al have agreed that the result of computational analysis is not fully reliable and can be skewed because of its limitations. Stewart et al agrees her team has to do future work to offer a better analysis since the indicators used can change. Brooks’ second article also shows that analysis of large data sets is good and can help with decisions but relying solely on computational analysis omits key aspects gained from traditional methods. Moreover, to rely solely on the traditional method is also not advisable as humans can incorporate their own feelings, culture, beliefs or prejudices that will skew the analysis. Both computational and traditional methods of analysis allow for a well-rounded result and a more accurately informed decision making process.
Works Cited
Brooks, David. “The Philosophy of Data”. New York Times 4 Feb. 2013: A23. Web. 1 Mar. 2013. <http://www.nytimes.com/2013/02/05/opinion/brooks-the-philosophy-of-data.html>
Brooks, David. “What Data Can’t Do”. New York Times 18 Feb. 2013: A23. Web. 1 Mar. 2013. < http://www.nytimes.com/2013/02/19/opinion/brooks-what-data-cant-do.html>
Lewis, Zamith and Hermida. “Content Analysis and Big Data”. Journal of Broadcasting & Electronic Media Mar. 2013: p34-p52. Web. 15 Mar 2013.
Stewart, Fisichella, Denecke. “Detecting Public Health Indicators from the Web for Epidemic Intelligence”. 2010. Web. 15 Mar 2013. <http://www.l3s.de/web/upload/documents/1/paper44_Stewart.pdf>
