search

lbb220 / GWmodel

GWmodel package
22 stars 13 forks source link

Multi scale GWR & use gwr as predictor #3

Open Frank-Fy opened 4 years ago

Frank-Fy commented 4 years ago

卢老师您好，我在学习您开发的这套工具包时遇到了一些问题，想向您咨询：

  1. 多尺度地理加权回归解决了在回归时不同自变量影响的空间范围不同的问题，但是在我在尝试过程中发现MGWR无法输出T统计和p值，请问有办法计算吗？因为我注意到蒙特卡洛模拟计算显著性的函数只能支持输入单一带宽，不适合多尺度（一组带宽）。
  2. gwr.predict函数可以用来预测缺失数据位置的系数和因变量，但它是基于标准GWR模型的，也就出现了和上面一样的问题，如何解决mgwr下的“一组带宽”的情况呢？

Hello, Mr. Lu, I have encountered some problems when learning the GWmodel you developed. I would like to consult you these:

  1. Multi-scale geographic weighted regression solves the problem that different independent variables affecting on different spatial scopes during regression. However, I found that MGWR could not output T statistics and p values.And I noticed that the monte-carlo simulation to caculate the Significant values can only support a single bandwidth, and is not suitable for multi-scale GWR (a list of bandwidth).
  2. gwr.predict function can be used to predict the coefficients and dependent variables of the location without data, but it is based on the basic GWR model, which leads to the same problem as above. How to solve the situation of "a list of bandwidth" under Multi-scale GWR?
lbb220 commented 4 years ago

谢谢关注！ From: Frank-Fy Date: 2019-11-22 02:33 To: lbb220/GWmodel CC: Subscribed Subject: [lbb220/GWmodel] Multi scale GWR & use gwr as predictor (#3) 卢老师您好，我在学习您开发的这套工具包时遇到了一些问题，想向您咨询： 多尺度地理加权回归解决了在回归时不同自变量影响的空间范围不同的问题，但是在我在尝试过程中发现MGWR无法输出T统计和p值，请问有办法计算吗？因为我注意到蒙特卡洛模拟计算显著性的函数只能支持输入单一带宽，不适合多尺度（一组带宽）。 目前MGWR支持计算T值，但其理论存在一定问题，未解决之前不能公开代码。 gwr.predict函数可以用来预测缺失数据位置的系数和因变量，但它是基于标准GWR模型的，也就出现了和上面一样的问题，如何解决mgwr下的“一组带宽”的情况呢？ MGWR目前不适合用于预测，因为其解决方案为back-fitting算法，与Basic GWR技术不同，无法在因变量未知的情形下进行模型解算。 Hello, Mr. Lu, I have encountered some problems when learning the GWmodel you developed. I would like to consult you these: Multi-scale geographic weighted regression solves the problem that different independent variables affecting on different spatial scopes during regression. However, I found that MGWR could not output T statistics and p values.And I noticed that the monte-carlo simulation to caculate the Significant values can only support a single bandwidth, and is not suitable for multi-scale GWR (a list of bandwidth). gwr.predict function can be used to predict the coefficients and dependent variables of the location without data, but it is based on the basic GWR model, which leads to the same problem as above. How to solve the situation of "a list of bandwidth" under Multi-scale GWR? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.