"The closer to the train station, the worse the kebab" – study

One programmer, equipped with Python and a bunch of software, went on a perilous pursuit to answer the age-old question, "Do kebabs get less tasty as you get closer to a train station?"

James Pae doesn’t actually like the conventional kebab. Nonetheless, he set out on a journey to determine whether kebabs are worse if they’re closer to the train station. A noble task that is not yet over.

After coming across this Reddit post asking anyone if they’re “motivated to prove” whether “the closer from the railway station, the less tasty the Kebab is.”

Anyone motivated to prove that? "the closer from the railway station the less tasty the Kebab is"
byu/Tight-Classroom4856 ingis

A wordy question, but pregnant with potential, Pae quickly took to the task, using various methods to extract all the data he needed to answer the age-old question.

As the original Reddit post was French, Pae decided to pick the rich city of Paris to conduct his experiment, a choice he might soon regret.

Pae told Cybernews that the initial statement, which was albeit re-worded for clarity, “implied a relationship between the walking distance from a train station and the quality of kebab-serving establishments.”

After applying this logic, he decided he needed to “measure this in some way.” Pae decided he needed three key data points:

• A network of pedestrian paths (sidewalks, crossings, tunnels, etc)
• Train (and Metro) station entrances
• Restaurants alongside a measure of quality (Google Places and their reviews)

Pae explained that with this data, he could “use the network to find the shortest path between any given restaurant and the series of station entrances.”

After this, he could “plot the relationship between the places’ review score and the distance to the nearest station entrance.”

To gather this data, Pae mainly used Python, including a variety of libraries, “notably OSMNx,” which allowed him to download the necessary network, station entrances, and Google Maps – “which let me conveniently make queries to the Google Places API.”

He also used NetworkX for network path-finding and GeoPandas for other areas of data processing.

“I also used matplotlib to generate the plots and QGIS for the screenshots and to inspect the data,” Pae told Cybernews.

With all of that work, while it only took less than a day, surely he’d have an accurate graph or map to see whether this hypothesis is true or false, right? Wrong.

The assumption that there was a linear relationship between the distance from the station and review rating was perhaps naive, and the location Pae chose, Paris, also presented some challenges.

“I've seen evidence that strengthens the hypothesis, especially in smaller French cities, though the relationship is far less obvious in Paris,” Pae said.

Furthermore, Google reviews “aren’t ideal,” as there could be many false reviews, among other factors that could’ve thrown the experiment off.

Nonetheless, Pae isn’t ready to give up just yet. Although Pae has “no clue” when he will start, but he wants to start again soon.

“The methodology for a follow-up absolutely needs to be changed. Instead of just Paris, I'll need to sample many smaller cities or suburbs. Also, the hypothesis will be switched from anything implying a linear relationship to something like "Is kebab the closest to a train station the worst?" (or something along those lines),” Pae concludes.

Did we figure out if the further away you are from the train station, the tastier the kebab? Well, no.

Pae’s findings were rather entropic and didn’t clearly demonstrate the hypothesis.

Many highly rated kebab shops were actually near the station, but similarly, mid or low-rated restaurants were also nearby.

The graph shows that the further away you got from the station, the fewer kebabs there were to find, which could also sway the results as perhaps people view these kebabs as better as they don’t have much choice around.

I fear that the question will never be answered. I guess we’ve got to wait for the second study.