我想知道是否有人能就我运行的负二项模型给出一些建议。基本上我试图通过各种分类变量和一些连续变量（例如权重1）来预测人数（数字）。下面的示例组成数据（省略一些变量）：
**Designation Habitat Year Weight1 Day DayBro Type_Day People**
SPA/SAC Heathland 2004 200 Tuesday NA Weekday Autumn 10
SPA/SAC Heathland 2004 450 Tuesday NA Weekday Autumn 0
SPA/SAC Heathland 2004 5000 Tuesday NA Weekday Autumn 7
SPA/SAC Heathland 2004 60 Tuesday NA Weekday Autumn 0
SSSI Heathland 2004 800 Sunday NA Weekend Autumn 6
SSSI Heathland 2004 3000 Sunday NA Weekend Autumn 9
SANG Heathland 2004 20 Saturday NA Weekend Autumn 50
SANG Heathland 2004 60 Saturday NA Weekend Autumn 3
SPA/SAC Heathland 2004 50 Wednesday NA Weekday Autumn 88
SPA/SAC Heathland 2004 50 Wednesday NA Weekday Autumn 0
SPA/SAC Heathland 2004 70 Wednesday NA Weekday Autumn 5
运行命令
model.nb = glm.nb(People2 ~ DayBro + Designation + Habitat + Type_Day +
Season + Designation + Weight1 + Weight2 + Weight3, data = TAA3)
我得到以下结果
Call:
glm.nb(formula = People2 ~ Weight1 + Weight2 + Weight3 + DayBro +
Designation + Habitat + Type_Day + Season + Designation,
data = TAA3, init.theta = 0.7571378169, link = log)
Deviance Residuals:
Min 1Q Median 3Q Max
-2.7536 -1.0111 -0.4665 0.1315 3.7172
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 3.883e+00 1.510e-01 25.720 < 2e-16 ***
Weight1 -2.581e-05 7.230e-06 -3.570 0.000357 ***
Weight2 3.057e-04 5.151e-05 5.935 2.94e-09 ***
Weight3 -6.328e-03 1.453e-03 -4.354 1.34e-05 ***
DayBroTerm -3.345e-01 9.597e-02 -3.486 0.000490 ***
DesignationSSSI 2.540e-01 2.183e-01 1.164 0.244573
DesignationSANG 2.533e-01 1.643e-01 1.542 0.123089
DesignationpSANG 5.618e-01 1.838e-01 3.056 0.002241 **
HabitatGrassland -7.616e-01 1.641e-01 -4.641 3.46e-06 ***
HabitatHeathland -4.467e-01 1.535e-01 -2.909 0.003624 **
HabitatMixed -3.555e-02 1.204e-01 -0.295 0.767751
HabitatWetland -3.569e-01 1.696e-01 -2.104 0.035348 *
HabitatWoodland -3.283e-01 2.642e-01 -1.242 0.214065
Type_DayWeekend 4.860e-01 7.802e-02 6.229 4.69e-10 ***
SeasonSummer 1.580e-01 1.353e-01 1.168 0.242926
SeasonAutumn 3.756e-01 1.520e-01 2.471 0.013482 *
SeasonWinter 1.332e-01 1.275e-01 1.045 0.295982
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for Negative Binomial(0.7571) family taken to be 1)
Null deviance: 1174.5 on 885 degrees of freedom
Residual deviance: 1049.3 on 869 degrees of freedom
(114 observations deleted due to missingness)
AIC: 8674
Number of Fisher Scoring iterations: 1
Theta: 0.7571
Std. Err.: 0.0332
2 x log-likelihood: -8637.9990
由于估算值和显着性值的差异，我对这些结果感到有些震惊。 Weight1（数值预测值）具有非常低的估计值且非常重要，而DesignationSSSI（分类预测器）具有更高的估计值且不重要。我知道，对于分类预测因子，它指的是基础类别，而对于Weight1，它表示Weight1的单位增加。
是否有人对此有任何意见/建议以及这些结果是否有意义？或者可能错误地指定了模型语法（等）中的某些内容？
非常感谢你的帮助！
达米亚诺