Poisson Regression

Spatial Statistics

Relative Risk for Poisson Regression

Yeongeun Jeon
12-16-2020

Poisson Regression


Relative Risk

실제로 \(y\)의 평균 보다 "Rate" or "Relative Risk"에 더 관심있으며, Poisson Regression의 목적은 Relative Risk을 추정하는 것이다. 왜냐하면 단순히 “수”로는 어느 게 크고 작은지 또는 높고 낮은지 정확하게 판단할 수 있는 기준이 없기 때문이다.

\[\begin{align*} y_{i} &\thicksim Poisson(\lambda_{i}), \;\; \lambda_{i} = E_{i}\rho_{i}\\ \eta_{i} &= \log{\rho_{i}}= \log{\frac{\lambda_{i}}{E_{i}}} = \beta_{0}+ \sum_{j} \beta_{j}x_{ji}\\ \beta_{j} &\thicksim Normal \end{align*}\]

Real Data

Loading Data

pacman::p_load("maptools",     # For readShapePoly
               "spdep",        # For poly2nb
               "dplyr", 
               "ggplot2",
               "RColorBrewer", # For brewer.pal
               "INLA")



dat.2019   <- read.csv("2019_crime.csv",header=T) 

# Convert rows in the order of ESPI_PK
dat.2019 <- dat.2019[order(dat.2019$ESRI_PK),]

head(dat.2019)
   ESRI_PK year district rape pop_total pop_femal sec_fac
10       0 2019   도봉구   90    335631    171670      10
12       1 2019   은평구  194    484546    251186      22
6        2 2019 동대문구  168    363023    184533      21
20       3 2019   동작구  251    408912    211206      18
18       4 2019   금천구  153    251820    122866      13
17       5 2019   구로구  226    439371    219769      17
   safe_return_use
10           12602
12           10145
6            20776
20           18766
18           17946
17           15233

Loading .shp

seoul.map   <- maptools::readShapePoly("./TL_SCCO_SIG_W_SHP/TL_SCCO_SIG_W.shp")   # Call .shp file
seoul.nb    <- poly2nb(seoul.map)      # Builds a neighbours list based on regions with contiguous boundaries
seoul.listw <- nb2listw(seoul.nb)      # Supplements a neighbours list with spatial weights for the chosen coding scheme
seoul.mat   <- nb2mat(seoul.nb)        # Generates a weights matrix for a neighbours list with spatial weights for the chosen coding scheme
                                       # Object of class "nb"

Mapping

Frequency

Relative Risk
or   <- dat.2019$rape
E    <- dat.2019$pop_total*( sum(dat.2019$rape) / sum(dat.2019$pop_total))
or.p <- dat.2019$rape/E

\[\begin{align*} \hat{\rho} = \frac{y_{i}}{E_{i}} = \frac{y_{i}/n_{i}}{\sum y_{i}/ \sum n_{i}} \end{align*}\]

Reuse

Text and figures are licensed under Creative Commons Attribution CC BY 4.0. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".