Notes

....................... 1 ........................................

We take a random sample of individuals in a population and
identify whether they smoke and if they have cancer.
We observe that there is a strong relationship between
whether a person in the sample smoked or not and whether
they have lung cancer.
We claim that the smoking is related to lung cancer in the larger population.
We explain we think that the reason for this relationship is because
cigarette smoke contains known carcinogens such as arsenic and benzene,
which make cells in the lungs become cancerous.

This is an example of an descriptive data analysis.

This is an example of a causal data analysis -----------------------> wrong

This is an example of an inferential data analysis. ------->

This is an example of a mechanistic data analysis.




..................... 2 .........................................................


What is the most important thing in Data Science?

Machine learning and prediction.

Knowing Hadoop and Pig.

Statistical inference.

Hacking skills.

The question you are trying to answer -----------------------------> correct


.......................... 3 ....................................................

If the goal of a study was to relate Martha Stewart
Living Subscribers to Our Site's Users based on the number of
people that lived in each region of the US, what would be the potential problem?


Source: http://xkcd.com/1138/

There would be confounding because the number of people that live
in an area is related to both Martha Stewart Living Subscribers and Our Site's Users. --->

We couldn't be sure whether subscribing to Martha Steward Living causes -----------------> wrong
people to be Users of Our Site or the other way around.

We would be performing inference on the relationship between Martha Stewart
Living Subscribers and Our Site's Users.

We wouldn't know the sensitivity of our predictions.


.......................... 4 ......................................

What is an experimental design tool that can be used to
address variables that may be confounders at the design phase of an experiment?

Fixing variables.

Data cleaning. ------------------> wrong

Using regression models.

Using data from a database.

Stratifying variables ----->

.......................... 5 ..................................

What is the reason behind the explosion of interest in big data?

The price and difficulty of collecting and storing data has dramatically dropped --->

There have been massive improvements in machine learning algorithms.

Technologies like Hadoop and Map Reduce started the big data era. ------> wrong

There have been massive improvements in statistical analysis techniques.