Difference between revisions of "ANLY482 AY2017-18 T2 Group 31 Model Buidling and Analysis"
Line 90: | Line 90: | ||
To determine: <br> | To determine: <br> | ||
1. To identify cluster of locations that have higher occurrence of indiscriminate parkings<br> | 1. To identify cluster of locations that have higher occurrence of indiscriminate parkings<br> | ||
+ | |||
+ | Function (kernel 𝑘) of a given radius (𝑟) “visits” each point in the study region. 𝑘 provides the weight of the area surrounding 𝑠 in proportion to its distance to 𝑠_𝑖 <br> | ||
+ | <div align="center"> | ||
+ | [[File:KDEformula.png|200px]]<br> | ||
+ | </div> | ||
+ | 𝑘 is calculated as a function of the distance between point 𝑠 and 𝑠_𝑖, over given radius 𝑟 <br> | ||
+ | The density of the study region is obtained by summing 𝑘 of all points 𝑠_𝑖 within 𝑟 <br> | ||
+ | <div align="center"> | ||
+ | [[File:LargeBW.png|400px]] [[File:SmallBW.png|400px]] | ||
+ | </div> | ||
+ | Kernel Density Estimations are sensitive to changes in radius values <br> | ||
+ | Large radius leads to a smoother curve, but local details would be obscured <br> | ||
+ | Small radius leads to many small spikes that are very localised <br> | ||
+ | Using the statistically significant radius distance obtained from Modified L Test as a search radius within each event<br> |
Revision as of 21:47, 14 April 2018
Model Building
Modified L Test via Ripley's K Function
To determine:
1. If the notifications appear to be clustered or randomly distributed in our area of interest
2. Minimum radius distance which shows signs of statistically significant clustering
Number of observed notifications is compared to the number of notifications expected based on Complete Spatial Randomness (CSR)
CSR assumes distribution of points is homogeneous over the study area
Null hypothesis: the spatial points are randomly distributed, using alpha = 0.01
Bold line represents the observed values for a range of 𝑟
Red dotted line represents the expected theoretical value for a range of 𝑟
Grey area is the confidence envelope obtained through 100 iterations of Monte Carlo procedures based on assumptions from CSR
For each simulated point pattern, 𝐾(𝑟) is estimated over a range of 𝑟. The max and min of these functions define an upper and lower simulation of the envelope
Converting K-function to L function, and to Modified L function
Interpreting the Modified L Test graph
Kernel Density Estimation
To determine:
1. To identify cluster of locations that have higher occurrence of indiscriminate parkings
Function (kernel 𝑘) of a given radius (𝑟) “visits” each point in the study region. 𝑘 provides the weight of the area surrounding 𝑠 in proportion to its distance to 𝑠_𝑖
𝑘 is calculated as a function of the distance between point 𝑠 and 𝑠_𝑖, over given radius 𝑟
The density of the study region is obtained by summing 𝑘 of all points 𝑠_𝑖 within 𝑟
Kernel Density Estimations are sensitive to changes in radius values
Large radius leads to a smoother curve, but local details would be obscured
Small radius leads to many small spikes that are very localised
Using the statistically significant radius distance obtained from Modified L Test as a search radius within each event