HW2 Clarification

HW2 Clarification

by Elan Guterman -

Hi,

I had a few questions about the homework. 

Questions 1-5: What do you mean when you ask to label our DAGs? Are you expecting that we use the A, Y, C, L, U notation and then include a key below the DAG, use the variables in the article, or choose whatever is easiest for our brain to handle?

Question 8b: I'm not clear about the direction of the relationship between firefighter and physical activity. The question stem seems to indicate both that good physical shape -> leads to being firefighter but additionally that firefighters -> have an affinity for physical activity. Did you want us to address this as time varying exposure? Or pick one particular element of this relationship?

Question 9: Are we supposed to assume that we have expert knowledge of the system to answer this question? And, if we are being asked to consider the statistical associations that may hint at possible confounding (rather than prove it because we don't have that expert knowledge), are there standard considerations for interpreting the size of the effect between the variables i.e. Pr[A = 1 | L = 1] / Pr[A = 1 | L = 0] and Pr[Y = 1 | L = 1] / Pr[Y = 1 | L = 0]? (Or, does this question suggest that I am approaching this question incorrectly?)

Hopefully this makes sense and my notation is not blatantly wrong !

Elan

Re: HW2 Clarification

by Rebecca Graff -

Hi Elan,

1. Choose whatever is easiest for your brain. Just be sure that we can tell what each node in the DAG represents.

2. Treat the relationship as physical activity ---> being a firefighter

3. The only expert knowledge you should need to answer the question is what is provided in the prompt. You can also assume that the numbers in the table are divided by one billion, such that any quantities you estimate from the data would have extremely narrow confidence intervals.

Best,

Rebecca

Re: HW2 Clarification

by Elan Guterman -
Thanks so much.

One additional point of clarification for the role of the variables in 8a - is the variable "sicker" meant to invoke CD4 count or be a separate measure? The question stem indicates how protective effect of ART on AIDS is mediated through CD4 count but that being sicker (?low CD4) leads to --> receiving ART.

Appreciate the guidance