Re: Survey Sampling Design

[Hans Chen]

Dear Lister
 
Numerous visitors and customers visit our business center for shopping and sightseeing each day and we have installed a camera system to record the times of the door open and door close (we call it door count). Unfortunately, this number is not exactly what we want due to following reason:
 
·         Numerous employees are working inside the business center. Once they enter and exit the building, they are counted, which make door count inflated.
·         When people enter the building in group, they are just counted as 1
·         ….
We know how many times the door have been opened/close each day, but we want to find out actually how many people enter/exit the building. Is that possible to conduct sampling to get a coefficient and adjust the dour count using the coefficient to get the number of victors, and how can we conduct such a sampling? Any suggestion will be highly appreciated.

Han Chen

[Boreak Silk]

Hi Han,

 

Two issues that you need to address - technology and estimate. We are doing a similar exercise in Melbourne but

in open space environment.

 

Technology - we tested various pedestrian counting sensors and the most accurate, that we could get a few years ago, are 

thermal/laser sensors installed under awnings/on public poles. Using these can address undercounting issue (many people

entering shoulder-to-shoulder but counted 1) issue and they have very high level of accuracy less than +/- 5 per cent in open 

space environment.

 

This leads to the second issue - estimate: to test the level of accuracy, we do manual count every 6-12 months. We look at

the hourly patterns (we have data down to minutely intervals) for different days (weekday, weekend day or special event day) 

and select an hour during peak hours, an hour during off peak hours for a typical weekday and typical weekend day, do manual

count and then compare with sensor count during at the same time. By doing a similar exercise, you probably can come up with

an adjustment factor for your data, for example, one count is equal to 1.5 people or something like that. You need to graph hourly

or half-hourly data for different days of the week and make your decision, based on your local knowledge, to drop the first and

last hour or half-an-hour counts from your analysis to account for employees working in the centre.

 

Another issue, at least with your system, is the same people might be counted more than once. I assume people might get in and out

of your centre more than once and you need to factor this in your analysis depending on how accurate you want the result to be.

 

 

Hope this helps.

 

Boreak

---

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Han Chen
Sent: Saturday, 22 October 2011 1:34 AM
To: [hidden email]
Subject: Re: Survey Sampling Design

Dear Lister
 
Numerous visitors and customers visit our business center for shopping and sightseeing each day and we have installed a camera system to record the times of the door open and door close (we call it door count). Unfortunately, this number is not exactly what we want due to following reason:
 
·         Numerous employees are working inside the business center. Once they enter and exit the building, they are counted, which make door count inflated.
·         When people enter the building in group, they are just counted as 1
·         ….
We know how many times the door have been opened/close each day, but we want to find out actually how many people enter/exit the building. Is that possible to conduct sampling to get a coefficient and adjust the dour count using the coefficient to get the number of victors, and how can we conduct such a sampling? Any suggestion will be highly appreciated.

Han Chen
This email is intended solely for the named addressee.
If you are not the addressee indicated please delete it immediately.