Exam related Q&A

Hi all,

Shall post answers to Qs I receive that I think can do with greater dissemination.

Update: Got these Qs from Amrita over email today. My responses are prefaced with '>>'

Hi Amrita,

Need some clarification on the following questions based on hw-session 8-

1. Pls explain the difference b/w standardized and unstandardized co-efficients? What are they used to interpret respectively?

>> Ordinarily, we cannot compare unstandardized betas because the scale of the variable affects the coeff. E.g., measuring weight in Kgs instead of in grams would raise the new weight coeff by 1000 times. Standardized coeffs are the coeffs we get when the X variables have been standardized (i.e. transformed to mean 0, variance 1). The coeffs can hence now be compared with one another.

2. For q4, hw session8, what do we refer to (which column of the coefficients table) to gague the impact on sales. Also, just an addition to point 2- usually we refer to the beat coefficients (and shall refer to t-statistic in absence of beta coefficients) and as per that the ranking should be- LnPrice, Bud, Miller, LnSize, Promo, Light. However, as per solution it is Bud, Miller, LnPrice, LnSize, Promo, Light which is according t-statistic. Hence need a clarification on this.

>>If standardized coeffs is given, we use it. If it is not given, we usually go with the (absolute) value of the t-stat. In the exam, your R output might not contain the standardized coeffs, then go with the t-stats. This was mentioned in a blog-post also. We look at the absolute value of the t-stat or th standardized betas. I used the absolute value (i.e. ignore sign) of the t-stats tobuild the list in q4, hw session 8.

3. For q1, session8, pls advise whether we need to refer to the ‘ standard coefficient- beta’ column or the ‘t-statistic’ column?

>> Either is fine. If std beta coeff is given ,go with it. I went with t-stat. I think much of this was explained in the blog posts mentioned in my last email to the class. Pls go thru those again carefully.

Sudhir

************************************

Shouri has sent in some more Qs. My responses to his earlier set of Qs is here.

My responses are prefaced with a '>>' symbol.

1. A scale say from Least likely (1) to Most likely (7) – this is a strictly an ordinal scale, right? Or can it pass of as an interval scale?

>>1-7 Likert is an interval scale. 'xyz Likely' is scale guidance. An interval scale has ordinal (direction info) properties, so yes, its ordinal too.

2. Please confirm that we are not covering the article “Forecasting the adoption of a new product”

>> We're not. More generally, *anything* not covered in class in both sections won't land up in the exam. I'd say go easy on the pre-reads as well. The in-class reads and guest lecture topics too are not important from the exam point of view, besides.

3. Is a perceptual map plotted on principal components or factors?

>> Principal components. More generally, whenever we want to "project" data into 2 dimensions for plotting or visualization purposes, we go with principal components. When we want to know about possible underlying constructs etc, we use common factor analysis (factor loadings table and all that).

4. Surrogate analysis has been mentioned as a qualitative way of demand estimation. Can you briefly explain what this is?

>> It is taking a good or service that exists as a surrogate for the proposed good or service and then trying to estimate demand for the new offering based on the evolution of demand for the identified surrogate. Thus, for example, we could say that 3D-printers will go the way of 2-D photo printing in its eventual adoption by the private sector - i.e. people can print photos at home but don't, they prefer photo kiosks for this; similarly, people could print objects at home in 3D but probably won't and will prefer kiosk model for this too. Etc.

I couldn't cover this in the second section, so its out of exam bounds. Again, "anything* in the slides that was not taken up in class is out of bounds.

Hope that clarifies. Shall append more Q&As as received to the top of this post.

Sudhir