HackerTrans
TopNewTrendsCommentsPastAskShowJobs

bobdosherman

no profile record

comments

bobdosherman
·letztes Jahr·discuss
Writing sharpens your thoughts. Working math problems sharpens your ability to do math. In the context of the education system though, where grades are a signal of future ability, there's a strong incentive to engage in rent-seeking by either searching for a solutions manual or "refining" an LLM's output. The little I've done when I've taught math-based econ is to make it clear that in-class tests have very high weight on your final grade, and out-of-class problem sets have very low weight. I can only mouth words as to why it's vital for students to struggle independently on the problem sets as a tool for learning.
bobdosherman
·vor 3 Jahren·discuss
There are way too many link-in-bio products out there that all do relatively the same thing
bobdosherman
·vor 3 Jahren·discuss
I used call point-and-click statistical software (like JMP) was the same as giving people who didn't know what they were doing a loaded gun. But democratizing access to advanced statistics...yada yada...who cares about asymptotic theory and identification and what not. Then R and Python and APIs that try to abstract as much as possible, and more loaded guns. But the talk of those loaded guns are really just phd-holders being obnoxious to some degree (but not completely wrong because stats can be misused...). But this really does seem like dumping a bunch of loaded guns all over the place. Nope
bobdosherman
·vor 5 Jahren·discuss
- Slide 38 has inconsistent signs for their cost of revenue line. Come on GS...you guys are left-lead. - They love to highlight their LTV/CAC ratio, so let me say that their choice of LTV definition is not ideal (footnote 1 on slide 19) - change in adj. gross profit is not necessarily just attributable to the cohort one acquires in a given year (since they're using S&M expense in a given year to represent CAC). This number could be easily biased upwards if previous cohorts are increasing their sub revenue in the current year (which is seems they are given the chart on slide 31). Assuming LTV of a customer is a perpetuity based on retention rates is an assumption many people make, but also will overstate the LTV. View this metric cautiously.
bobdosherman
·vor 5 Jahren·discuss
There also seems to be confusion that ESPN+ is equivalent to ESPN, but ESPN+ does not carry the linear programming of ESPN. Using part of the $15 YouTube TV discount to purchase ESPN+ will not make you whole on the lose of ESPN - still no live ESPN content.
bobdosherman
·vor 5 Jahren·discuss
If I estimate a model to predict an anomaly using data that never has any realized anomalies, how well will that model do out-of-sample? While framing this as a bad unrestricted ROE maximization problem is a nice simplification, it's not clear that having everyone move to a restricted ROE maximization subject to keeping assets large enough to insure against some unforecastable shock is welfare enhancing. That could be a lot of wasted insurance.

I will give him credit for cleverly spinning this logic all into a pitch to kill the unrealized cap gains tax proposal!
bobdosherman
·vor 5 Jahren·discuss
A lot of this comes down to preferences, and my preference ordering would be:

Ads that attempt to take me all the way from browsing-->discovery-->potential purchase-->purchase I view as always bad. Leave me alone while I'm browsing. Ads that attempt to take me from discovery-->potential purchase I view as ok but ineffective. At least throw a promo code in that ad next time, please and thank you. Ads that successfully take me from discovery-->potential purchase-->purchase I view as ok since my utility is higher having purchased the product given the payload of the ad (and I wouldn't have purchased the product having not gotten the ad).

One thing then from my (not necessarily everyones) preferences is that displaying an ad to me should only potentially be done if the conditional probability of me being in discovery mode is higher than the conditional probability of me being in browsing mode. And then it comes down to what of me is in that conditioning set. Some very fuzzy anonymous slice of me, well ok. But better not be PI in there...

But these are just my preferences, and they may not be representative. So there's an aggregation problem also. In general I'd also prefer if brands shifted the marginal marketing dollar towards channels where the disutility of showing me an ad when they estimate I'm in discovery mode but I'm actually in browsing mode is lowest - so (in my mind) when possible put more in influencer marketing versus google display network for instance. Ideally this preference is evident in brand's roi calcs so it's internalized.
bobdosherman
·vor 6 Jahren·discuss
Could also cross-subsidize by being a sub-affiliate network as part of an affiliate network. Company earns percentage of affiliate commissions produced by in-network links, which subsidize the non-commissionable out-of-network links (and non-earning in-network links).
bobdosherman
·vor 6 Jahren·discuss
Linear regression does not require errors to be iid, normal, and homoscedastic for it to "work". One of the ways to separate candidates is to push on what (and how) assumptions can be weakened, what the consequences are for estimation and inference, and what sort of corrections can be incorporated for maintaining consistency, improving efficiency, correcting biases, etc. An entry level candidate may not have (nor need to have) a complete understanding of asymptotic theory, but they should know what the purpose of robust standard errors are and how to use them.