遊蕩犬為疾病傳播源之一，且會與其他物種包括人類發生衝突。因此從生態平衡、公共衛生安全等不同角度來看，控制遊蕩犬的數量是一個重要的議題，且了解其分布後更可以幫助相關單位制訂遊蕩犬相關政策。全國性的遊蕩犬數量調查已於臺灣推行多年，隨著調查的進行，抽樣、實際執行與估計方式等也都會根據以往經驗進行調整，現在是一年兩次的全國範圍調查，採用不同層內設計的分層抽樣。

關於遊蕩犬總數的估計，以往使用設計基底（design-based）的估算方式推算各縣市遊蕩犬總數。然該估計方式無法估算未抽樣的鄉鎮市區和村里的遊蕩犬數量，進而無法詳細分析遊蕩犬分布情形。而本論文中欲根據模型基底（model-based）之估算方式對資料進行配適，進而預測未抽樣村里中的遊蕩犬數量，不僅可更加細緻描述遊蕩犬的分布，更可繪製遊蕩犬分布地圖。在本論文中，模型的選擇是根據均方根法（Root-Sum-Squares）和 AIC 值評估了幾種模型的配適與預測表現。其中使用的模型包含普通對數克里金法（Ordinary log-normal Kriging）、無隨機效應的負二項模型（Negative binomial model）與具有隨機效應的負二項模型（Negative binomial mixed effect models with spatial random effects model）。配適結果提供各縣及全臺遊蕩犬估計數量及信賴區間。此外，還可以通過所使用的模型預測未抽樣村里的遊蕩犬數量呈現相關的分布圖藉以預測可能的「熱點」區域，以利制定更合適的遊蕩犬相關政策。

Roaming dogs could spread diseases, and lead to the conflict between themselves and other animals including humans, therefore, it is an important issue to control the roaming dog population from different perspectives including ecological balance, public safety and health. Understanding the number and distribution structure of the roaming dogs is essential knowledge for the related agency to formulate proper and efficient policies to monitor the roaming dog population. The survey of the number of roaming dogs has been practiced in Taiwan for years. This survey has been modified several times based on the previous experience of execution including the survey period and sampling design, and now it is a biannual nationalwide survey in which a stratified sampling with different within-strata designs is used.

The usual design-based estimation method has been used in this survey to estimate the total number of roaming dogs in each county of Taiwan. Nevertheless, the number of roaming dogs cannot be estimated with the design-based estimation in the townships/districts and villages not included in the sample. Therefore, the detailed roaming dog distribution structure cannot be realized. However, the number of roaming dogs in the unsampled villages can be predicted based on the model-based inference, hence, the distribution structure can possibly be described on a map. Several model-based estimation methods are evaluated, based on the root-sum-of-squares and AIC values in this research. The model-based inference being considered includes Ordinary log-normal Kriging and negative binomial models with or without
spatial random effects. The estimated number of roaming dogs in each county and the whole Taiwan area are provided with associated confidence intervals, so that the population size of the roaming dogs can be monitored. In addition, the number of roaming dogs of the unsampled villages can be predicted as well with the model-based methods, and the associated distribution map can be presented to indicate the “hot-spot” of the roaming dogs, and more appropriate roaming dog population control policy can be formulated accordingly.

[1] 政府資料開放平臺. https://data.gov.tw/, 2020.
[2] Hirotugu Akaike. A new look at the statistical model identification. IEEE transactions on automatic control, 19(6):716–723, 1974.
[3] Ac´a¸cio Cardoso Amaral, Michael P Ward, and Joana da Costa Freitas. Estimation of roaming dog populations in timor leste. Preventive veterinary medicine, 113(4):608–613, 2014.
[4] Luc Anselin. Local indicators of spatial association—lisa. Geographical analysis,
27(2):93–115, 1995.
[5] Margaret Armstrong and Azeddine Boufassa. Comparing the robustness of ordinary kriging and lognormal kriging: Outlier resistance. Mathematical Geology, 20(4):447–457, 1988.
[6] Vin´ıcius Silva Belo, Guilherme Loureiro Werneck, Eduardo Sergio da Silva, David Soeiro Barbosa, and Claudio Jose Struchiner. Population estimation methods for free-ranging dogs: a systematic review. PLoS One, 10(12):e0144830, 2015.
[7] K Bogel, Karl Frucht, George Drysdale, Jenny Remfry, World Health Organization, et al. Guidelines for dog population management. Technical report, World Health Organization, 1990.
[8] Jianguo Chen, Lin Liu, Luzi Xiao, Chong Xu, and Dongping Long. Integrative analysis of spatial heterogeneity and overdispersion of crime with a geographically weighted negative binomial model. ISPRS International Journal of GeoInformation, 9(1):60, 2020.
[9] Executive Yuan Council of Agriculture. Council of agriculture, executive yuan, r.o.c.(taiwan). Council of Agriculture, Executive Yuan, R.O.C.(Taiwan), 2017.
[10] Noel Cressie. Spatial prediction and ordinary kriging. Mathematical geology, 20(4):405–421, 1988.
[11] Noel Cressie. Statistics for spatial data. John Wiley & Sons, 2015.
[12] Ricardo Augusto Dias, Aline Gil Alves Guilloux, Mauro Riegert Borba, Maria Cristina de Lourdes Guarnieri, Ricardo Prist, Fernando Ferreira, Marcos Amaku, Jos´e Soares Ferreira Neto, and Mark Stevenson. Size and spatial distribution of stray dog population in the university of s˜ao paulo campus, brazil. Preventive Veterinary Medicine, 110(2):263–273, 2013.
[13] Ver´onica Flores, Gustavo Viozzi, Carlos Rauque, Guillermo Mujica, Eduardo Herrero, Sebasti´an A Ballari, Luciano Ritossa, Gabriela Miori, Gilda Garibotti, Daniela G Zacharias, et al. A cross-sectional study of free-roaming dogs in a patagonian city: Their distribution and intestinal helminths in relation to socioeconomic aspects of neighborhoods. Veterinary Parasitology: Regional Studies and Reports, 33:100747, 2022.
[14] World Society for the Protection of Animals. Surveying roaming dog populations: Guidelines on methodology. 2007.
[15] Matthew E Gompper. The dog-human-wildlife interface: assessing the scope of the problem. Free-ranging dogs and wildlife conservation, pages 9–54, 2014.
[16] Yuying Hsu, Lucia Liu Severinghaus, and James A Serpell. Dog keeping in taiwan: its contribution to the problem of free-roaming dogs. Journal of Applied Animal Welfare Science, 6(1):1–23, 2003.
[17] Joelene Hughes and David W Macdonald. A review of the interactions between free-roaming domestic dogs and wildlife. Biological Conservation, 157:341–351, 2013.
[18] Daniel G Krige. A statistical approach to some basic mine valuation problems on the witwatersrand. Journal of the Southern African Institute of Mining and Metallurgy, 52(6):119–139, 1951.
[19] Frank J Massey Jr. The kolmogorov-smirnov test for goodness of fit. Journal of the American statistical Association, 46(253):68–78, 1951.
[20] Mark N Maunder and Andr´e E Punt. Standardizing catch and effort data: a review of recent approaches. Fisheries research, 70(2-3):141–159, 2004.
[21] Michelle K Morters, Trevelyan J McKinley, Olivier Restif, Andrew JK Conlan, Sarah Cleaveland, Katie Hampson, Helen R Whay, I Made Damriyasa, and James LN Wood. The demography of free-roaming dog populations and applications to disease and population control. Journal of Applied Ecology, 51(4):1096–1106, 2014.
[22] Fran¸cois Rousset. An introduction to the spamm package for mixed models. 2017.
[23] Tenzin Tenzin, Rubaiya Ahmed, Nitish C Debnath, Garba Ahmed, and Mat Yamage. Free-roaming dog population estimation and status of the dog population management and rabies control program in dhaka city, bangladesh. PLoS neglected tropical diseases, 9(5):e0003784, 2015.
[24] Jorge Kazuo Yamamoto. On unbiased backtransform of lognormal kriging estimates. Computational Geosciences, 11(3):219–234, 2007.
[25] Shih-Ching Yen, Yu-Ten Ju, Pei-Jen Lee Shaner, and Hsiang Ling Chen. Spatial and temporal relationship between native mammals and free-roaming dogs in a protected area surrounded by a metropolis. Scientific reports, 9(1):1–9, 2019.
[26] Achim Zeileis, Christian Kleiber, and Simon Jackman. Regression models for count data in r. Journal of statistical software, 27(8):1–25, 2008.
[27] 呂芷晴. 官方證實無誤！犬貓正式名列外來入侵種 - 窩窩｜專注為動物發聲的獨立媒體. 窩窩wuowuo, 2022.