5 ms·

The big one that's missing: There is nothing you can conclude from your data. It's clean, it makes its way properly to the analyst, and yet, there's just nothin

by d--b 10y ago

The big one that's missing: There is nothing you can conclude from your data. It's clean, it makes its way properly to the analyst, and yet, there's just nothing there...

searine 10y ago
Hypothesis driven data science is over the head of too many people.
- TeMPOraL 10y ago
  That's why it's called "data mining" - you keep digging until you find something. And with lots of data, you can always find something if you look hard enough - which leads to things like http://tylervigen.com/discover http://tylervigen.com/discover.
  - searine 10y ago
    We always called it a fishing expedition. Though, that's what has always appealed to me about truly large data sets, low risk of turing up jack-squat.
  - blahi 10y ago
    Data drudging is what it is called.
martingoodson 10y ago
A great example of this was reported to me from a head of data science at a (UK) national newspaper. The business set a task of predicting subscriber numbers using online user behaviour from the previous month as features. But it turned out that most subscribers converted within three days of first hitting the site. The previous month of data was almost completely worthless. Like you say, the data just didn't contain the information that it 'should' have.
eanzenberg 10y ago
If done properly that's still a useful result. "None of the variables were predictive of X."
- hobofan 10y ago
  And it's much less useful cousin hiding in the shadows "The variables are actually predictive of X, but you don't have enough data that it shows". (The chance for which isn't that big when done "properly".)
  - huac 10y ago
    At least in online contexts, this often means 'there might be an effect, but it is smaller than our experiment could have predicted. Let's keep adding samples.'
    - angry_octet 10y ago
      '... until we get the answer we're expecting'. Just expecting an effect is a bias towards outliers. http://doingbayesiandataanalysis.blogspot.com.au/2013/11/optional-stopping-in-data-collection-p.html?m=1 http://doingbayesiandataanalysis.blogspot.com.au/2013/11/opt...

by d--b 10y ago

The big one that's missing: There is nothing you can conclude from your data. It's clean, it makes its way properly to the analyst, and yet, there's just nothing there...

searine 10y ago

Hypothesis driven data science is over the head of too many people.

TeMPOraL 10y ago
That's why it's called "data mining" - you keep digging until you find something. And with lots of data, you can always find something if you look hard enough - which leads to things like http://tylervigen.com/discover http://tylervigen.com/discover.
- searine 10y ago
  We always called it a fishing expedition. Though, that's what has always appealed to me about truly large data sets, low risk of turing up jack-squat.
- blahi 10y ago
  Data drudging is what it is called.

martingoodson 10y ago

A great example of this was reported to me from a head of data science at a (UK) national newspaper. The business set a task of predicting subscriber numbers using online user behaviour from the previous month as features. But it turned out that most subscribers converted within three days of first hitting the site. The previous month of data was almost completely worthless. Like you say, the data just didn't contain the information that it 'should' have.

eanzenberg 10y ago

If done properly that's still a useful result. "None of the variables were predictive of X."

hobofan 10y ago
And it's much less useful cousin hiding in the shadows "The variables are actually predictive of X, but you don't have enough data that it shows". (The chance for which isn't that big when done "properly".)
- huac 10y ago
  At least in online contexts, this often means 'there might be an effect, but it is smaller than our experiment could have predicted. Let's keep adding samples.'
  - angry_octet 10y ago
    '... until we get the answer we're expecting'. Just expecting an effect is a bias towards outliers. http://doingbayesiandataanalysis.blogspot.com.au/2013/11/optional-stopping-in-data-collection-p.html?m=1 http://doingbayesiandataanalysis.blogspot.com.au/2013/11/opt...