Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Analyze how often users look on both sides of the road as an indicator of user accuracy #36

Open
daotyl000 opened this issue Aug 1, 2019 · 13 comments
Open

Analyze how often users look on both sides of the road as an indicator of user accuracy #36

daotyl000 opened this issue Aug 1, 2019 · 13 comments
Labels
enhancement New feature or request New Feature New features we can log using interaction or labels data

Comments

@daotyl000
Copy link
Contributor

How much does a user's accuracy change by how much they look around horizontally, and whether or not they look at both sides of the road? Heading records the position the user is looking horizontally, the value itself isn't helpful but we can use the range. Would we want to have the full 360 degrees or would a lower degrees be sufficient.
@daotyl000
Copy link
Contributor Author

This graph shows the average heading range of each users
Blue = Good users
Black = Bad users
Red = Users in neighborhoods without sidewalks

Screen Shot 2019-08-01 at 10 56 57 AM

All users (except for one bad users) who on average looked at more than about 140 degrees of the pano were good. This indicates that users who generally look more of the pano will have a higher accuracy. However, I am not sure what degrees range average would be sufficient, only one user had an averge over 225. While there are alot of good users who also had low ranges, the split where almost no bad users had a heading range of over about 120 degree.

@daotyl000
Copy link
Contributor Author

daotyl000 commented Aug 1, 2019 •
edited
Loading

When looking at number of times where a users looked at atleast 350 degrees of a pano, we see a similar trend where almost all users who had more than 300 panos where they viewed most of it were good users. I used 350 degrees because it is most of the pano and when I had used 360 degrees, all users collectively only totaled about 6 panos

Screen Shot 2019-08-01 at 11 18 11 AM

@jonfroehlich
Copy link
Member

Can you clearly define your x-axis. In other words: What is 'average heading' and how is it calculated? Similarly, what is 'number of 350+ degree views' and how is it calculated?

Also, can you start to apply a best fit linear regression line and report the fit value?

@daotyl000
Copy link
Contributor Author

daotyl000 commented Aug 1, 2019 •
edited
Loading

Average heading of on average, how many degrees does a user look around a pano. Heading is a value that is attatched to each action, I believe 0 is directly north.

number of 350+ degrees views means, the number of panos where the user had panned around to view atleast 350 degrees of the pano. This is calculated by substracting the lowest heading value by the highest heading value

What do you mean by the fit value? Does that mean the slope?

@jonfroehlich
Copy link
Member

jonfroehlich commented Aug 1, 2019 via email

@daotyl000
Copy link
Contributor Author

Here is the graph calculating the average range of degrees that the user looks at per pano:

Screen Shot 2019-08-01 at 3 45 57 PM

The correlation coefficient and significance are both 0.23

Here is the graph counting the number of times a user looks at the majority of a pano:

Screen Shot 2019-08-01 at 3 44 32 PM
The correlation coefficient is 0.4 while the significance value is 0.03

@jonfroehlich
Copy link
Member

jonfroehlich commented Aug 1, 2019 via email

@daotyl000
Copy link
Contributor Author

According to the scipy website (p-value = significance) : "The p-value roughly indicates the probability of an uncorrelated system producing datasets that have a Pearson correlation at least as extreme as the one computed from these datasets. The p-values are not entirely reliable but are probably reasonable for datasets lager than 500 or so."
I wasn't sure if this is what you had wanted.

@jonfroehlich
Copy link
Member

jonfroehlich commented Aug 1, 2019 via email

@daotyl000
Copy link
Contributor Author

daotyl000 commented Aug 1, 2019 •
edited
Loading

No so it can be ignored. The command to display the correlation value also displayed that value so I left it in.

@jonfroehlich
Copy link
Member

jonfroehlich commented Aug 2, 2019 via email

@daotyl000
Copy link
Contributor Author

Like this?

Screen Shot 2019-08-02 at 9 46 32 AM
Screen Shot 2019-08-02 at 10 01 41 AM

@jonfroehlich
Copy link
Member

jonfroehlich commented Aug 2, 2019 via email

@daotyl000 daotyl000 added the enhancement New feature or request label Aug 2, 2019
@daotyl000 daotyl000 added the New Feature New features we can log using interaction or labels data label Aug 28, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement New feature or request New Feature New features we can log using interaction or labels data
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@jonfroehlich @daotyl000