Hi all,
I've recently found myself doing the same exercise as another poster on this
thread to find potentially bogus entries - averaging postcodes per region
and finding ones that are significantly outside where they belong.
This raises a question though, let's say I believe these codes are suspect
and so does everyone else (indeed I've checked one or two on other websites
which use more "official" postcode data), should these codes be removed or
at least marked somehow as bogus?
If I'm going to use this data in a project of my own I could remove these
entries having manually checked them, but then the data correction isn't
being done in a shared way along with the data collection.
What's the best way of setting a feedback mechanism like this up? If
necessary I don't mind helping out?
Russ
--
View this message in context: http://www.nabble.com/Marking---flagging-incorrect-entries-tp14233507s26798…
Sent from the NPEmap - Talk mailing list archive at Nabble.com.
npemap,
I have compiled a list of entries which vary greatly from their
outbound's standard deviation: (worst entries at the top)
http://code.firefishy.com/files/bad-postcodes.csv
I am no statistician. Please correct me if I am off the mark.
/ Grant
------------------------
SQL:
SELECT `postcodes`.`outward`, `postcodes`.`inward`, `postcodes`.`lat`,
`postcodes`.`long`,`outbound_stats`.*,
abs(`postcodes`.`lat`-`outbound_stats`.`lat_avg`) AS `lat_offset`,
abs(`postcodes`.`long`-`outbound_stats`.`long_avg`) AS `long_offset`
FROM `postcodes` LEFT JOIN `outbound_stats` ON `postcodes`.`outward` =
`outbound_stats`.`outward` WHERE `outbound_stats`.`number` >= 10 AND (
(abs(`postcodes`.`lat`-`outbound_stats`.`lat_avg`) +
abs(`postcodes`.`long`-`outbound_stats`.`long_avg`)) >
(`outbound_stats`.`lat_stddev`*3 + `outbound_stats`.`long_stddev`*3 ))
ORDER BY (`lat_offset`+`lat_offset`) DESC
You may have noted a spate of bad code reports north of Brighton from me last
evening. As a new postcode mapper of - oh, about 3 minutes - I've been
checking the data by using the postcode map at free the postcode:
http://dev.openstreetmap.org/~random/postcodes/
This uses the npe codes as well as free the postcode codes to produce a
postcode map at, for exapmle, BN1 level. Bad codes show as unlikely or even
crazy post code zones on the map. For example, islands of BN1 in BN5 zones
and even of SP1 in BN. These seem to be bad postcodes on the NPEmap and
I've reported them.
Probably a well known method but useful. let me know if there is anything I
can do.
Best wishes
Mark
--
View this message in context: http://www.nabble.com/checking-for-bad-postcodes-tf4930997s26798.html#a1411…
Sent from the NPEmap - Talk mailing list archive at Nabble.com.