Posted by: qmaxim | January 31, 2014

Facebook shutting down?

Recently, researchers at Princeton, one of the premier US institutions of higher learning   after a thorough study concluded that if the present trends continues, Facebook will cease to exist by 2017. This is  contrary to recent trends. Facebook is steadily growing users,  reported record recent quarter and replaced yahoo as the 2nd largest internet advertisement revenue earner.

Are the days when people (particularly youngsters)  constantly updating their Facebook pages & checking their Facebook news feed at all hours of day and night  & at all places finally  over? Is Mark Zukerberg the CEO of Facebook, having sleepless nights about his billions disappearing? He is worth US$ 32.9  billion at the moment.

Well, it looks unlikely at least at present, even though  its popularity among very young seems to  be declining slightly. In a hilarious rebuttal,  by using similar approach  Facebook researchers have  predicted  that Princeton university will   shut down by 2021. And air we breathe will run out by 2060 and mankind as we know of course will be no more.

What is happening here?

The amount of data generated has been increasing exponentially  in the last few years. Along with the data deluge, software for analysis (many of them free, such as R) are becoming readily available and used by many. Most  of the enormous amount of data so  generated daily is in    public domain.  Data from Twitter feeds, NASA, US govt, US weather service, Google search data, etc is freely available . Along with the data explosion  there is boom in the number data crunchers called data scientists who work on interpreting this data to gain insights, unearth relationships and importantly to make predictions.

By using these powerful  easily available data analysis tools it is possible to unearth relationship between many variables (sometimes in 1000s)  and the variable of interest. One commonly used measure of strength of the relationship (or  correlation ) is the R2, generally  higher the  R2 stronger is the relationship. In the financial/ marketing  fields this could be as high as .7 even .9,  whereas in the drug discovery  .05 is considered to be a decent number. Machine learning computer algorithms like Neural Networks can model complex relationships even when they  are not obvious or even when do not seem to be plausible.

This is where the problem arises. High correlation number is does not imply that one causes the other. This  has to be proved by logic, theory or prior knowledge. This is where even experienced scientists sometimes tip over. This seems to be the case in this context. Princeton researchers in their study modeled spread of  social networks on spread of contagious diseases. On seeing steady decline of  web search for the word ‘Facebook’ they concluded that Facebook will cease to exist by 2017. The reason for this finding was  quite simple; most users have Facebook app on their phones and have no need to search for this key word anymore.
This is the classic case of  equating  correlation with causation.

One famous  example to explain this concept is the  following story.  Statistical study  of shark attacks at a certain beach indicated that there is good correlation between amount of ice cream sold and number of shark attacks. So, it was concluded that eating large amounts of ice cream causes  increased shark attacks.

The reality was more mundane. During some summer months due to  dwindling food in their traditional hunting grounds sharks come closer to beach looking for food. During the same time of the year,  number of  sun bathers and  surfers thronging  the beach  exploded. They became easy target for sharks looking for food.   And during hot  summer months sunbathers consume lots  of ice cream.

Another similar case  study is the near perfect correlation (R2=.97) between fresh lemons imported from Mexico into US and total US highway fatality rate.

I will be pleased to hear your views

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: