Пример оценки многомерной логит-модели. Ученые собрали данные по 672 знакомствам М и Ж в ночном клубе (348 М и 672 Ж) - из книги Discovering Statistics Using R. Объясняемая переменная определяет следующие варианты дальнейшего развития событий:
Разговоры имеют следующие характеристики:
Предыдущие исследование указывают на то, что результативность характеристик зависит от пола собеседника, то есть interactions
library(mlogit)## Loading required package: Formula
## Loading required package: maxLik
## Loading required package: miscTools
##
## Please cite the 'maxLik' package as:
## Henningsen, Arne and Toomet, Ott (2011). maxLik: A package for maximum likelihood estimation in R. Computational Statistics 26(3), 443-458. DOI 10.1007/s00180-010-0217-1.
##
## If you have questions, suggestions, or comments regarding the 'maxLik' package, please use a forum or 'tracker' at maxLik's R-Forge site:
## https://r-forge.r-project.org/projects/maxlik/chatData <- read.delim('datasets/Chat-Up Lines.dat', header = TRUE)
head(chatData)## Success Funny Sex Good_Mate Gender
## 1 Get Phone Number 3 7 6 Male
## 2 Go Home with Person 5 7 2 Male
## 3 Get Phone Number 4 6 6 Male
## 4 Go Home with Person 3 7 5 Male
## 5 Get Phone Number 5 1 6 Male
## 6 Get Phone Number 4 7 5 MalechatData$Gender <- relevel(chatData$Gender, ref = 2)
mlChat <- mlogit.data(chatData, choice = "Success", shape = 'wide') Расчет модели
##
## Call:
## mlogit(formula = Success ~ 1 | Good_Mate + Funny + Gender + Sex +
## Gender:Sex + Funny:Gender, data = mlChat, reflevel = 3, method = "nr",
## print.level = 0)
##
## Frequencies of alternatives:
## No response/Walk Off Get Phone Number Go Home with Person
## 0.39216 0.47549 0.13235
##
## nr method
## 6 iterations, 0h:0m:0s
## g'(-H)^-1g = 0.00121
## successive function values within tolerance limits
##
## Coefficients :
## Estimate Std. Error t-value
## Get Phone Number:(intercept) -1.783070 0.669772 -2.6622
## Go Home with Person:(intercept) -4.286354 0.941398 -4.5532
## Get Phone Number:Good_Mate 0.131840 0.053726 2.4539
## Go Home with Person:Good_Mate 0.130019 0.083521 1.5567
## Get Phone Number:Funny 0.139389 0.110126 1.2657
## Go Home with Person:Funny 0.318456 0.125302 2.5415
## Get Phone Number:GenderFemale -1.646223 0.796247 -2.0675
## Go Home with Person:GenderFemale -5.626369 1.328589 -4.2348
## Get Phone Number:Sex 0.276206 0.089197 3.0966
## Go Home with Person:Sex 0.417283 0.122083 3.4180
## Get Phone Number:GenderFemale:Sex -0.348326 0.105875 -3.2900
## Go Home with Person:GenderFemale:Sex -0.476639 0.163434 -2.9164
## Get Phone Number:Funny:GenderFemale 0.492441 0.139992 3.5176
## Go Home with Person:Funny:GenderFemale 1.172404 0.199240 5.8844
## Pr(>|t|)
## Get Phone Number:(intercept) 0.0077631 **
## Go Home with Person:(intercept) 5.284e-06 ***
## Get Phone Number:Good_Mate 0.0141306 *
## Go Home with Person:Good_Mate 0.1195351
## Get Phone Number:Funny 0.2056135
## Go Home with Person:Funny 0.0110376 *
## Get Phone Number:GenderFemale 0.0386891 *
## Go Home with Person:GenderFemale 2.287e-05 ***
## Get Phone Number:Sex 0.0019577 **
## Go Home with Person:Sex 0.0006307 ***
## Get Phone Number:GenderFemale:Sex 0.0010020 **
## Go Home with Person:GenderFemale:Sex 0.0035409 **
## Get Phone Number:Funny:GenderFemale 0.0004354 ***
## Go Home with Person:Funny:GenderFemale 3.996e-09 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Log-Likelihood: -868.74
## McFadden R^2: 0.13816
## Likelihood ratio test : chisq = 278.52 (p.value = < 2.22e-16)## exp.chatModel.coefficients.
## Get Phone Number:(intercept) 0.16812128
## Go Home with Person:(intercept) 0.01375498
## Get Phone Number:Good_Mate 1.14092570
## Go Home with Person:Good_Mate 1.13885057
## Get Phone Number:Funny 1.14957104
## Go Home with Person:Funny 1.37500360
## Get Phone Number:GenderFemale 0.19277659
## Go Home with Person:GenderFemale 0.00360163
## Get Phone Number:Sex 1.31811957
## Go Home with Person:Sex 1.51783194
## Get Phone Number:GenderFemale:Sex 0.70586855
## Go Home with Person:GenderFemale:Sex 0.62086652
## Get Phone Number:Funny:GenderFemale 1.63630634
## Go Home with Person:Funny:GenderFemale 3.22974620Показать таблицу результатов модели в виде для публикаций
## Version: 1.35
## Date: 2015-04-25
## Author: Philip Leifeld (University of Konstanz)
##
## Please cite the JSS article in your publications -- see citation("texreg").| Model 1 | |
|---|---|
| Get Phone Number:(intercept) | -1.78** |
| (0.67) | |
| Go Home with Person:(intercept) | -4.29*** |
| (0.94) | |
| Get Phone Number:Good_Mate | 0.13* |
| (0.05) | |
| Go Home with Person:Good_Mate | 0.13 |
| (0.08) | |
| Get Phone Number:Funny | 0.14 |
| (0.11) | |
| Go Home with Person:Funny | 0.32* |
| (0.13) | |
| Get Phone Number:GenderFemale | -1.65* |
| (0.80) | |
| Go Home with Person:GenderFemale | -5.63*** |
| (1.33) | |
| Get Phone Number:Sex | 0.28** |
| (0.09) | |
| Go Home with Person:Sex | 0.42*** |
| (0.12) | |
| Get Phone Number:GenderFemale:Sex | -0.35** |
| (0.11) | |
| Go Home with Person:GenderFemale:Sex | -0.48** |
| (0.16) | |
| Get Phone Number:Funny:GenderFemale | 0.49*** |
| (0.14) | |
| Go Home with Person:Funny:GenderFemale | 1.17*** |
| (0.20) | |
| AIC | 1765.47 |
| Log Likelihood | -868.74 |
| Num. obs. | 1020 |
| p < 0.001, p < 0.01, p < 0.05 | |