Kyla Scanlon  

@scanlon_kyla

interests include writing, yoga, and math // rts not endorsements; views are my own

Vrijeme pridruživanja: srpanj 2014.

Tweetovi

Blokirali ste korisnika/cu @scanlon_kyla

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @scanlon_kyla

  1. Prikvačeni tweet
    22. sij

    Do you own a car and drive less than 10,000 miles a year? You're wasting money, according to my analysis.

    Prikaži ovu nit
    Poništi
  2. Here's to new datasets and new code and new models! (Also as a PSA Google just released their dataset search tool. It's decent) END

    Prikaži ovu nit
    Poništi
  3. It was an attempt to model video game preorders as a form of currency and compare potential ROIs to btc, USD, gld (I know, I know). It was meant to be a fun project but I ended up data dredging, which is never ideal. There's value in starting again and freeing up brain space.

    Prikaži ovu nit
    Poništi
  4. Today I had to throw out an article I'd been working on for a while. The data were good. But the model wasn't. I was spinning my wheels. I will release a new article soon. One that doesn't require p-hacking to model a relationship that doesn't make sense.

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    30. sij

    Tech Twitter TLDR is out: 🏛 Founders Fund takes on local gov via 🇩🇪 Germany's startup problem via ✨ Magic of ads via and + tweets from , , , , , etc.

    Poništi
  6. 30. sij

    I've been following Michael & Ben for years. Huge honor to be featured on their latest podcast! Give it a listen, if you haven't already!

    Poništi
  7. 22. sij

    **Disclaimer: none of this is investment advice, and I have no affiliation with any ride-sharing company

    Prikaži ovu nit
    Poništi
  8. 22. sij

    20/ Overall, the driving experience is pretty subjective. From a quantitative viewpoint, it’s sometimes cheaper to get a ride share. Qualitatively, it all depends on what you value.

    Prikaži ovu nit
    Poništi
  9. 22. sij

    19/ EXTREMELY IMPORTANT This data doesn’t consider the changes that are coming with California AB5, which could radically change all of the model

    Prikaži ovu nit
    Poništi
  10. 22. sij

    18/ There are a lot of reasons that the data could be telling this story, including how I priced cars and my personal ride-sharing data. However, ride-sharing is a better option for low-mileage users as compared to driving according to this analysis.

    Prikaži ovu nit
    Poništi
  11. 22. sij

    17/ In California, it makes zero sense to own a car, according to these numbers. (I think this is a bit extreme, but I would imagine that it is more cost effective to use rideshare in a big city vs rural America)

    Prikaži ovu nit
    Poništi
  12. 22. sij

    16/ I’ve color-coded each to show the point where you would be better off driving. For most of the models, the cross-over point is between 9k –12k miles. Hybrid, electric, sedan: car cost > rideshare cost up to 9k annual miles. SUV: car > rideshare up to 12k miles

    Prikaži ovu nit
    Poništi
  13. 22. sij

    15/ I held seconds constant in the analysis for the first two models and included rush hour in the output. For the U.S. analysis, I set indirect costs equal to $3,000 to tease out the cost of California living.

    Prikaži ovu nit
    Poništi
  14. 22. sij

    14.5/ (i know it's text output. i have no excuses)

    Prikaži ovu nit
    Poništi
  15. 22. sij

    14/ Compare Models The R-2 doesn’t vary too much between models despite the change in variables. Unfortunately, the R-2isn’t very high :( The addition of more variables and more data would improve that. My sample size was small and I only had three true variables.

    Prikaži ovu nit
    Poništi
  16. 22. sij

    13/ Model 3: Drop seconds, interaction term between dist and rush hour IT: assumption that the effect of distance on fare is different for rush hour vs non-rush hour Drive 1 mile at 5 pm = $30 4 am = $3 Proxy for surge pricing.

    Prikaži ovu nit
    Poništi
  17. 22. sij

    12/ Model 2: Including a Dummy Variable for Rush Hour Rush hour: bt 6am — 9am and 3pm — 7pm (1) Any time outside of that window is non-rush hour (0) I had 40 rides that were outside of rush hour and 18 that were within rush hour specifications.

    Prikaži ovu nit
    Poništi
  18. 22. sij

    11/ multicollinearity - high correlation between two variables. we can't have our predictors predicting each other. I address this by dropping seconds and creating an instrumental variable (Model 3)

    Prikaži ovu nit
    Poništi
  19. 22. sij

    10/ Model 1: Seconds and Distance I ran a basic linear regression in R to see the predictive power of seconds & distance in determining fare, since that is what most ride-sharing models are based on HOWEVER 91% correlation bt seconds and distance = multicollinearity problem

    Prikaži ovu nit
    Poništi
  20. 22. sij

    9/ To illustrate the issue of extrapolation, consider what Robert Chira once said about Apple’s valuation: “If you extrapolate far enough out into the future, to sustain that growth Apple would have to sell an iPhone to every man, woman, child, animal and rock on the planet.

    Prikaži ovu nit
    Poništi
  21. 22. sij

    8/ This is a correlation plot between my 3 main rideshare variables, fare, seconds, and distance. The goal of this analysis is to predict the fare of the rideshares. I will be extrapolating my data out into the thousands of miles.

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·