3 ms·

I'm the author of this post and the founder of dataguru.hk, AMA

by slygent 8y ago

baxtr 8y ago
Nice work! Do you see any differences when slicing the city into its various districts?
- slygent 8y ago
  Great idea, I'll add it to the to-do pile!
aequitas 8y ago
There is no significant drop for 40 where 34 and 44 both do have these. Is this because 40 is pronounced differently and doesn't have the connotation?
- slygent 8y ago
  Compare 39 to 40, and 49 to 50... I think it's just harder for people to avoid ALL floors starting with 4, but there is still a small effect.
diamondo25 8y ago
Its unclear whether the amount of transactions is compensated for the floors that do not exist (thus less supply).
- slygent 8y ago
  But before that I include a graph of flats available, and indeed there are fewer flats with floor numbers ending with 4! So an interesting analysis would be to see if the turnover (transactions divided by availability) is different... But I think price paid is more interesting :)
JorgeGT 8y ago
Very interesting work! Just a quick note, for this kind of evenly spaced, discrete X data, stairs plots are usually better. For instance this highlights better the 13th floor issue, which in the line plot was obscured: https://i.imgur.com/cSjT5xx.png https://i.imgur.com/cSjT5xx.png
- slygent 8y ago
  As an R aficionado, I totally agree, but it's not available on Metabase yet, which is the front-end I want to offer non-technical clients to visualise data. I should put in a PR...
- neic 8y ago
  I agree that a stepped line graph is better that a line graph, but since the data is discrete and not continuous a lollipop chart or maybe a bar chart would convey it better. Monty from xiph.org learned me that: https://xiph.org/video/vid2.shtml https://xiph.org/video/vid2.shtml @ 5:57
  - JorgeGT 8y ago
    Also a good option, although they seem a bit "busy" for my taste when Y values are very close: https://i.imgur.com/4ePNXbp.png https://i.imgur.com/4ePNXbp.png
ksec 8y ago
What happen to Dataguru.hk between 2013 and 2018?

I'm the author of this post and the founder of dataguru.hk, AMA

baxtr 8y ago

Nice work! Do you see any differences when slicing the city into its various districts?

slygent 8y ago
Great idea, I'll add it to the to-do pile!

aequitas 8y ago

There is no significant drop for 40 where 34 and 44 both do have these. Is this because 40 is pronounced differently and doesn't have the connotation?

slygent 8y ago
Compare 39 to 40, and 49 to 50... I think it's just harder for people to avoid ALL floors starting with 4, but there is still a small effect.

diamondo25 8y ago

Its unclear whether the amount of transactions is compensated for the floors that do not exist (thus less supply).

slygent 8y ago
But before that I include a graph of flats available, and indeed there are fewer flats with floor numbers ending with 4! So an interesting analysis would be to see if the turnover (transactions divided by availability) is different... But I think price paid is more interesting :)

JorgeGT 8y ago

Very interesting work! Just a quick note, for this kind of evenly spaced, discrete X data, stairs plots are usually better. For instance this highlights better the 13th floor issue, which in the line plot was obscured: https://i.imgur.com/cSjT5xx.png https://i.imgur.com/cSjT5xx.png

slygent 8y ago
As an R aficionado, I totally agree, but it's not available on Metabase yet, which is the front-end I want to offer non-technical clients to visualise data. I should put in a PR...
neic 8y ago
I agree that a stepped line graph is better that a line graph, but since the data is discrete and not continuous a lollipop chart or maybe a bar chart would convey it better. Monty from xiph.org learned me that: https://xiph.org/video/vid2.shtml https://xiph.org/video/vid2.shtml @ 5:57
- JorgeGT 8y ago
  Also a good option, although they seem a bit "busy" for my taste when Y values are very close: https://i.imgur.com/4ePNXbp.png https://i.imgur.com/4ePNXbp.png

ksec 8y ago

What happen to Dataguru.hk between 2013 and 2018?