Session 3 go done y'day.
It was a 'dry and technical' session, admittedly, owing to the nature of subject matter.... but, it cannot be denied that we (both you and me) did try to enliven things a bit here and there, didn't we?
1. We covered sampling basics - notational and definitional stuff included. Now, with sampling done, in theory, you are equipped enough to design a full fledged survey based primary data collection survey exercise... only.
-------------------------------------------------
2. Significantly, y'day also marked the first use of R in DC. The data and code are put up as .txt files on LMS. YOu are encouraged to pls try replicating my classroom results at home. Of course, replications won't be perfect because of the very nature of 'random' sampling but that's OK.
-------------------------------------------------
3. We also delved into the business experimentation space. This is a rapidly evolving space and one, I believe, that defines the frontiers in demand-side analytics.
If you were to ask me which is a promising area within analytics to build skill-sets in, I'd promptly say 'Experimentation analytics'.
-------------------------------------------------
4. We studied 'traditional' experimental design in Session 3. Traditionally, experiments were used to measure the *average* treatment effect across the sample. This was particularly true in the natural sciences. However, increasingly in business, we find that the average is misleading and not good enough.
In a whole host of modern businesses (both web based and brick-and-mortar), the treatment effect of interest is produced by exposing a micro-segment (in extreme cases, a segment of One) to a causal condition (in extreme cases, a product or service custom-defined for that micro segment) and measuring the outcome difference from the average for the market as a whole.
It is this combination of product design and micro-segmentation that gives the new age business experimentation its edge and its own distinctive flavour. It forms the subject of the last reading in session 3.
I believe I did not get to emphasise this point enough, particularly in Section A. I strongly encourage people to read the relevant 2007 HBR article on business experimentation provided in your course-pack.
-------------------------------------------------
5. By the way, this particular trend I refer to in point 4 above isn't necessarily restricted to the business sector either.... Politics is not immune. For instance, take this Washington Post article from June 2013 that gives a layman's introduction to how Sri Obama leveraged Big data and microtargetting techniques for his 2012 campaign.
Here is a short 5 minute video that makes a similar point. And here is a more detailed, longish article from the MIT technology review that goes into more detail. I have every reason to believe that at least some of these ideas will or have found their way into India's coming MahaBhArat - the general elections of 2014....
-------------------------------------------------
6. One last bit about traditional experiments before I sign off. Here's an interesting article on a piece of academic research that aimed to test (using traditional 'True' experimental design) whether social networks make us smarter or dumber...
Seeking to find out if social networks make us smarter a team of scientists investigated if networks help us imitate analytical thought processes from our peers.To carry out their experiment the researchers tested university students with a series of brain-straining questions. 100 volunteers were separated into 5 social networks each with 20 individuals. Connections between the people in the networks were assigned randomly by a computer to fit 5 different network patterns. At one extreme all the people in the network were connected directly to all the others, and at the other extreme there were no connections at all. To test how these networks helped the people in them to learn, the scientists quizzed the volunteers with a 'cognitive reflection test', a series of questions which rely on analytical reasoning to overcome incorrect intuition.
To see if the social networks helped the people in them to improve their answers the volunteers were asked each of the questions 5 times. The first time the volunteers had to figure it out on their own, the next 5 times they were allowed to copy the answer from their neighbours in the network. The researchers found that in well connected networks ...
OK, so what did they find? Well, to find out I suggest you read the entire article only....
-------------------------------------------------
OK, that's it from me for now. See you in class today for an R joy ride into the Text analytics skies...
Sudhir
No comments:
Post a Comment