Poster

[sahirbhatnagar]

Can you put Dr. Abbass Khalili on 1 line? so that it doesnt cut his name in half

[sahirbhatnagar]

Measures of economic performance include GDP, unemployment rate, and stock indices; these measures were summarized using their first two principal components. A GLM is used to test the association between economy and TV time use, with time being included smoothly.

I think we should speak in the same tense

[sahirbhatnagar]

We report that watching television is associated with unemployment rate, and that the association changes with gender (but not with region, race or income).

This is ignoring or implying that tv time is not associated with stock indices...I realise that result of mine came in very late today, but I think we should include that also:

We report that watching TV and TV usage are associated with both PC's, and that the association changes with gender (but not with region, race or income)

[sahirbhatnagar]

Economy and Stratification

title should be more descriptive, maybe :

_Principal Components of Economic Measures and Stratification _

By the way, is the y axis of the stratification plot a count? it looks like a proportion

[sahirbhatnagar]

The section header should be Methods (plural).

i think the $\alpha$ in the linear predictor should have a prime on it, so that we're clear that its a vector. this will differentiate it further from the fact that beta1 and beta2 are scalars

is it confusing that the rate parameter of the gamma is $\lambda$ and the parameter for the smoothing is also lambda?

[sahirbhatnagar]

Limitation: We used PCA to derive the economic measures used in the model, which ignores the relationship with the response

IMO, this is a stronger limitation than the auto regressive nature of the beta econ.

[sahirbhatnagar]

Its not clear to me what the difference between the section: Selected socio-demographic variables

and below in the results section you write: "the selected factors are"

it sounds like you're mentioning the same information twice

[kevinmcgregor]

I think in the main challenges section there should be a question mark at the end of "How can economy be measured".

[turgeonmaxime]

Thanks for the comments, I have made the corrections you mentionned. To answer some the questions raised above:

• the y-axis is indeed a proportion (proportion of counts coming from a specific category);
• I don't think the two uses of lambda is confusing, but to be on the safe side I changed the tuning parameter to tau;
• I agree about the limitation you raised when it comes to PCA, but I don't think it's major. The original question was "how to we define economy?", which does not depend on TV use. PCA was used to define economic variables. On the other hand, I think the auto-regressive nature is a major issue. It's as if we used least squares instead of weighted least squares, and so I think we end up underestimating our standard errors/confidence intervals;
• Kevin: would you agree then that we should put question marks everywhere in the main challenges section? (except of course the first bullet point)

[kevinmcgregor]

Yes, I agree with your final statement.