More project related guidance

Class,

Met with another group just now. Answered their questions and would like to share with the wider class what advice I had to give.

IMO, the basic problem in the project - at its heart - is a demand estimation exercise. This problem, BTW, is at the heart of what econometrics tries to solve.

At its core, the demand function can be written as

sales (Y) = f (price, other attributes, preferences over attributes, consumer
profiles etc)

The sales and price data come best from secondary sources. Long back, Sreenath had emailed the whole class the list of units sold for each brand-model. Price data are available from carazoo.com. Other attribute data about the cars are similarly available.

Well, XYZ could have done the above secondary data analysis itself, why hire you for the job? Turns out the primary dataset you have has critical information that goes into the recommendation.

Its not enough to say, in response to Q1 that "Yes, enter" and then for Q2a) "Enter hatchbacks". Say also, why enter hatchbacks, who the target clientele is likely to be, what their preferences are in terms of attributes, what their current satisfaction and dissatisfaction levels are, what media they watch and hence how best they can be reached, what demographic and psychographic attributes best enable identifying this target segment, why they are the most attractive/vulnerable target segment from XYZ's POV etc etc. You get the drift.


So folks, rather than have you waste time on fruitless pursuits such as cleaning up all the data columns we have etc, better to refocus your energies on the core problem.

Also, pls make a simple to read flowchart version of your research design and include that in your slides. That is your methodology section centerpiece in your deliverable report. A simple flowchart showing how you will arrive at the answer, what info you will need at each step will force clarity in thinking and planning. One suggestion - Start with the research questions at the top and break it down step by step into manageable parts for which data are needed. Simplifies this process of planning and implementing, IMHO.


Chalo, thats enough write-ups for today.

Am now working furiously on getting R to run conditional logits so that you will have the code with you should you need logit for the project down the line. SPSS has choked on running the more complex logit models. R and SAS can do it (I've used them previously for logit analyses), heck even excel can do logit variants.

Jai Ho.

Sudhir