sqtab: sqtab: models for square tables
In catspec: Special models for categorical variables
Description
Usage
Arguments
Details
Value
Author(s)
References
See Also
Examples
Description
These functions are used to estimate loglinear models for square tables such as quasi-independence, quasi-symmetry. Examples show how these models can be incorporated in a multinomial logistic model with covariates.
Usage
1
2
3
4
5
6
7
8
mob.qi(rowvar, colvar, constrained = FALSE, print.labels = FALSE)
mob.eqmain(rowvar, colvar, print.labels = FALSE)
mob.symint(rowvar, colvar, print.labels = FALSE)
mob.cp(rowvar,colvar)
mob.unif(rowvar, colvar)
mob.rc1(rowvar, colvar, equal = FALSE, print.labels = FALSE)
fitmacro(object)
check.square(rowvar, colvar, equal = TRUE)
Arguments
rowvar
Factor representing the row variable
colvar
Factor representing the column variable
print.labels
If FALSE (default) then numeric values rather
than factor values are printed for compact results
equal
(mob.rc1) If TRUE, a homogeneous row and column effects model 1 with equal scale values for the row and column variables is estimated. Otherwise, a (regular) RC1 model is estimated with different scale values for the row and column variables
(check.square) If TRUE, the row and column variables must have the same number of categories
constrained
(mob.qi)If TRUE, a quasi-independence model-constrained is estimated with a single parameter for the diagonal cells
object
(fitmacro) An object of class glm for family=poisson and link=log
Details
These functions are used to estimate loglinear models for square tables:
- mob.qi
Quasi-independence
- mob.eqmain
Equal main effects (Hope's halfway model)
- mob.symint
Symmetric interaction
- mob.cp
Crossings-parameter model
- mob.unif
Uniform association
- mob.rc1
Row and columns model 1
- fitmacro
Calculates BIC and AIC relative to a
saturated loglinear model
- check.square
Internal function to check if row and column variables are both factors with the same number of levels
These functions create part of the design matrix for a loglinear model. With the exception of mob.eqmain the models are an interaction effect between the row and column variable in order to test for a certain pattern of association. As with most things in R, these functions represent one way of doing things.
These restricted models can also be applied in a multinomial logistic regression model. In order to do this, the data have a “long” shape. Rather than one record per subject, the dataset must have ncat records per subject, where ncat is the number of categories of the dependent variable in the multinomial logistic model. The dataset must also have variables indexing the choices and indicating the choice made by each subject.
The documentation for clogit has an example that shows how the data can be restructured in this way. The mlogit.data function can also restructure the data as required. The models can be subsequently estimated using clogit or mlogit.data.
Value
mob.qi
A factor that will produce coefficients for the diagonal cells of a table, using off- diagonal cells as base category
mob.eqmain
A design matrix with equality contraints on the main effects
mob.symint
A design matrix for an interaction with equality contraints on coefficients on opposite sides of the diagonal
mob.cp
A set of vectors for a crossings-parameter model
mob.unif
A vector for a uniform association model
mob.rc1
A set of vectors for a row and columns model 1
fitmacro
Prints deviance, df, BIC, AIC, number of parameters and N
check.square
Stops function if either the row or column variable is not a factor or if the number of levels is unequal
Author(s)
John Hendrickx <John_Hendrickx@yahoo.com>
References
Agresti, Alan. (1990). Categorical data analysis. New York: John Wiley & Sons.
Allison, Paul D. and Nicholas Christakis. (1994). Logit models for sets of ranked items. Pp. 199-228 in Peter V. Marsden (ed.), Sociological Methodology. Oxford: Basil Blackwell.
Breen, Richard. (1994). Individual Level Models for Mobility Tables and Other Cross- Classifications. Sociological Methods & Research 33: 147-173.
Goodman, Leo A. (1984). The analysis of cross-classified data having ordered categories. Cambridge, Mass.: Harvard University Press.
Hendrickx, John. (2000). Special restrictions in multinomial logistic regression. Stata Technical Bulletin 56: 18-26.
Hendrickx, John, Ganzeboom, Harry B.G. (1998). Occupational Status Attainment in the Netherlands, 1920-1990. A Multinomial Logistic Analysis. European Sociological Review 14: 387-403.
Hout, Michael. (1983). Mobility Tables. Sage Publication 07-031.
Logan, John A. (1983). A Multivariate Model for Mobility Tables. American Journal of Sociology 89: 324-349.
See Also
glm,
[mlogit]mlogit,
[mlogit]mlogit.data,
[survival]clogit, [survival]coxph, [nnet]multinom
Examples
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
# Examples of loglinear models for square tables,
# from Hout, M. (1983). "Mobility Tables". Sage Publication 07-031
# Table from page 11 of "Mobility Tables"
# Original source: Featherman D.L., R.M. Hauser. (1978) "Opportunity and Change."
# New York: Academic, page 49
Freq <- c(
1414, 521, 302, 643, 40,
724, 524, 254, 703, 48,
798, 648, 856, 1676, 108,
756, 914, 771, 3325, 237,
409, 357, 441, 1611, 1832)
OccFather<-gl(5,5,labels=c("Upper nonmanual","Lower nonmanual","Upper manual","Lower manual","Farm"))
OccSon<-gl(5,1,labels=c("Upper nonmanual","Lower nonmanual","Upper manual","Lower manual","Farm"))
FHtab <- data.frame(OccFather,OccSon,Freq)
xtabs(Freq~OccFather+OccSon,data=FHtab)
# independence model
indep<-glm(Freq~OccFather+OccSon,family=poisson(),data=FHtab)
summary(indep)
fitmacro(indep)
wt <- as.numeric(OccFather != OccSon)
qi0<-glm(Freq~OccFather+OccSon,weights=wt,family=poisson(),data=FHtab)
# A quasi-independence loglinear model, using structural zeros
# (page 23 of "Mobility Tables").
# 0 1 1 1 1 values of variable "wt"
# 1 0 1 1 1
# 1 1 0 1 1
# 1 1 1 0 1
# 1 1 1 1 0
qi0<-glm(Freq~OccFather+OccSon,weights=wt,family=poisson(),data=FHtab)
summary(qi0)
fitmacro(qi0)
# Quasi-independence using a "dummy factor" to create the design
# vectors for the diagonal cells (page 23).
# 1 0 0 0 0
# 0 2 0 0 0
# 0 0 3 0 0
# 0 0 0 4 0
# 0 0 0 0 5
glm.qi<-glm(Freq~OccFather+OccSon+mob.qi(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.qi)
fitmacro(glm.qi)
# Quasi-independence without using the functions
# Factor labels prevent numeric comparisons, create numeric versions
# of the row and column variables
OccFather_n <- unclass(OccFather)
OccSon_n <- unclass(OccSon)
q1 <- ifelse(OccFather_n==OccSon_n & OccSon_n==1,1,0)
q2 <- ifelse(OccFather_n==OccSon_n & OccSon_n==2,1,0)
q3 <- ifelse(OccFather_n==OccSon_n & OccSon_n==3,1,0)
q4 <- ifelse(OccFather_n==OccSon_n & OccSon_n==4,1,0)
q5 <- ifelse(OccFather_n==OccSon_n & OccSon_n==5,1,0)
glm.qi2<-glm(Freq~OccFather+OccSon+q1+q2+q3+q4+q5,family=poisson(),data=FHtab)
summary(glm.qi2)
fitmacro(glm.qi2)
# Quasi-independence constrained (QPM-C, page 31)
# Single immobility parameter
# 1 0 0 0 0
# 0 1 0 0 0
# 0 0 1 0 0
# 0 0 0 1 0
# 0 0 0 0 1
glm.q0<-glm(Freq~OccFather+OccSon+mob.qi(OccFather,OccSon,constrained=TRUE),family=poisson(),data=FHtab)
# slightly different results than Hout also found in Stata: L2=2567.658, q0=0.964
summary(glm.q0)
fitmacro(glm.q0)
# Quasi-symmetry using the symmetric cross-classification (page 23)
# 0 1 2 3 4 values of variable "sym"
# 1 0 5 6 7
# 2 5 0 8 9
# 3 6 8 0 10
# 4 7 9 10 0 */
glm.qsym<-
glm(Freq~OccFather+OccSon+mob.symint(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.qsym)
fitmacro(glm.qsym)
symmetry<-glm(Freq~mob.eqmain(OccFather,OccSon)
+mob.symint(OccFather,OccSon),family=poisson(),data=FHtab)
summary(symmetry)
fitmacro(symmetry)
# Crossings parameter model (page 35)
# 0 v1 v1 v1 v1 | 0 0 v2 v2 v2 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | 0 0 v2 v2 v2 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | v3 v3 v3 0 0 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | v3 v3 v3 0 0 | v4 v4 v4 v4 0
glm.cp<-glm(Freq~OccFather+OccSon+mob.cp(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.cp)
fitmacro(glm.cp)
# Uniform association model: linear by linear association (page 58)
glm.unif<-glm(Freq~OccFather+OccSon+mob.unif(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.unif)
fitmacro(glm.unif)
# RC model 1 (unequal row and column effects, page 58)
# Fits a uniform association parameter and row and column effect
# parameters. Row and column effect parameters have the
# restriction that the first and last categories are zero.
glm.rc1<-glm(Freq~OccFather+OccSon+mob.rc1(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.rc1)
fitmacro(glm.rc1)
# Homogeneous row and column effects model 1 (page 58)
# An equality restriction is placed on the row and column effects
glm.hrc1<-glm(Freq~OccFather+OccSon+mob.rc1(OccFather,OccSon,equal=TRUE),family=poisson(),data=FHtab)
# Results differ from those in Hout, replicated by other programs
summary(glm.hrc1)
fitmacro(glm.hrc1)
#-------------------------------------------------------------------------------
# Examples on using these models in multinomial logistic regression
#-------------------------------------------------------------------------------
# Data from the 1972-78 GSS used by Logan (1983)
library(survival)
data(logan)
# Restructure the data in 'long' format using mlogit.data
library(mlogit)
pc <- mlogit.data(logan, shape = "wide", choice = "occupation")
head(pc,10)
# pc$alt indexes choices, pc$chid indexes respondents
# pc$occupation is a boolean variable that is TRUE where pc$alt corresponds
# with the actual choice made
class(logan$occupation) #"factor"
class(pc$alt) #"character"
# pc$alt needs to be transformed into a factor for use in the some models
# for square tables which assume ordered categories
pc$alt <- factor(pc$alt,levels=c("farm", "operatives", "craftsmen", "sales", "professional"))
# A regular multinomial logit model
m1<-mlogit(occupation ~ 0 | education+race, data=pc, reflevel = "farm")
summary(m1)
# Estimate a "quasi-uniform association" loglinear model for "focc" and "occupation"
# with "education" and "race" as covariates at the respondent level
# The quasi-uniform association model contains alternative-specific effects,
# education and race are individual specific effects
m2 <- mlogit(occupation ~ mob.qi(focc,alt)+mob.unif(focc,alt) | education+race,data=pc, reflevel = "farm")
summary(m2)
# Same model using survival::clogit
# First create a design matrix for the alternatives that drops the first category
occ.X<-model.matrix(~pc$alt)[,-1]
head(occ.X,10)
# The clogit output exceeds the default 80 characters
options(width=256)
# pc$chid which indexes respondents must be used in strata()
# Individual specific effecs are modelled as interactions between the
# design matrix for pc$alt and the covariates
cl.qu<-clogit(occupation~occ.X+occ.X:education+occ.X:race+
mob.qi(focc,alt)+mob.unif(focc,alt)+strata(chid),data=pc)
summary(cl.qu)
catspec documentation built on May 1, 2019, 8:21 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Description Usage Arguments Details Value Author(s) References See Also Examples
Description
These functions are used to estimate loglinear models for square tables such as quasi-independence, quasi-symmetry. Examples show how these models can be incorporated in a multinomial logistic model with covariates.
Usage
1 2 3 4 5 6 7 8 | mob.qi(rowvar, colvar, constrained = FALSE, print.labels = FALSE)
mob.eqmain(rowvar, colvar, print.labels = FALSE)
mob.symint(rowvar, colvar, print.labels = FALSE)
mob.cp(rowvar,colvar)
mob.unif(rowvar, colvar)
mob.rc1(rowvar, colvar, equal = FALSE, print.labels = FALSE)
fitmacro(object)
check.square(rowvar, colvar, equal = TRUE)
|
Arguments
rowvar |
Factor representing the row variable |
colvar |
Factor representing the column variable |
print.labels |
If FALSE (default) then numeric values rather than factor values are printed for compact results |
equal |
( |
constrained |
( |
object |
( |
Details
These functions are used to estimate loglinear models for square tables:
- mob.qi
Quasi-independence
- mob.eqmain
Equal main effects (Hope's halfway model)
- mob.symint
Symmetric interaction
- mob.cp
Crossings-parameter model
- mob.unif
Uniform association
- mob.rc1
Row and columns model 1
- fitmacro
Calculates
BICandAICrelative to a saturated loglinear model- check.square
Internal function to check if row and column variables are both factors with the same number of levels
These functions create part of the design matrix for a loglinear model. With the exception of mob.eqmain the models are an interaction effect between the row and column variable in order to test for a certain pattern of association. As with most things in R, these functions represent one way of doing things.
These restricted models can also be applied in a multinomial logistic regression model. In order to do this, the data have a “long” shape. Rather than one record per subject, the dataset must have ncat records per subject, where ncat is the number of categories of the dependent variable in the multinomial logistic model. The dataset must also have variables indexing the choices and indicating the choice made by each subject.
The documentation for clogit has an example that shows how the data can be restructured in this way. The mlogit.data function can also restructure the data as required. The models can be subsequently estimated using clogit or mlogit.data.
Value
mob.qi |
A factor that will produce coefficients for the diagonal cells of a table, using off- diagonal cells as base category |
mob.eqmain |
A design matrix with equality contraints on the main effects |
mob.symint |
A design matrix for an interaction with equality contraints on coefficients on opposite sides of the diagonal |
mob.cp |
A set of vectors for a crossings-parameter model |
mob.unif |
A vector for a uniform association model |
mob.rc1 |
A set of vectors for a row and columns model 1 |
fitmacro |
Prints deviance, df, BIC, AIC, number of parameters and N |
check.square |
Stops function if either the row or column variable is not a factor or if the number of levels is unequal |
Author(s)
John Hendrickx <John_Hendrickx@yahoo.com>
References
Agresti, Alan. (1990). Categorical data analysis. New York: John Wiley & Sons.
Allison, Paul D. and Nicholas Christakis. (1994). Logit models for sets of ranked items. Pp. 199-228 in Peter V. Marsden (ed.), Sociological Methodology. Oxford: Basil Blackwell.
Breen, Richard. (1994). Individual Level Models for Mobility Tables and Other Cross- Classifications. Sociological Methods & Research 33: 147-173.
Goodman, Leo A. (1984). The analysis of cross-classified data having ordered categories. Cambridge, Mass.: Harvard University Press.
Hendrickx, John. (2000). Special restrictions in multinomial logistic regression. Stata Technical Bulletin 56: 18-26.
Hendrickx, John, Ganzeboom, Harry B.G. (1998). Occupational Status Attainment in the Netherlands, 1920-1990. A Multinomial Logistic Analysis. European Sociological Review 14: 387-403.
Hout, Michael. (1983). Mobility Tables. Sage Publication 07-031.
Logan, John A. (1983). A Multivariate Model for Mobility Tables. American Journal of Sociology 89: 324-349.
See Also
glm,
[mlogit]mlogit,
[mlogit]mlogit.data,
[survival]clogit, [survival]coxph, [nnet]multinom
Examples
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 | # Examples of loglinear models for square tables,
# from Hout, M. (1983). "Mobility Tables". Sage Publication 07-031
# Table from page 11 of "Mobility Tables"
# Original source: Featherman D.L., R.M. Hauser. (1978) "Opportunity and Change."
# New York: Academic, page 49
Freq <- c(
1414, 521, 302, 643, 40,
724, 524, 254, 703, 48,
798, 648, 856, 1676, 108,
756, 914, 771, 3325, 237,
409, 357, 441, 1611, 1832)
OccFather<-gl(5,5,labels=c("Upper nonmanual","Lower nonmanual","Upper manual","Lower manual","Farm"))
OccSon<-gl(5,1,labels=c("Upper nonmanual","Lower nonmanual","Upper manual","Lower manual","Farm"))
FHtab <- data.frame(OccFather,OccSon,Freq)
xtabs(Freq~OccFather+OccSon,data=FHtab)
# independence model
indep<-glm(Freq~OccFather+OccSon,family=poisson(),data=FHtab)
summary(indep)
fitmacro(indep)
wt <- as.numeric(OccFather != OccSon)
qi0<-glm(Freq~OccFather+OccSon,weights=wt,family=poisson(),data=FHtab)
# A quasi-independence loglinear model, using structural zeros
# (page 23 of "Mobility Tables").
# 0 1 1 1 1 values of variable "wt"
# 1 0 1 1 1
# 1 1 0 1 1
# 1 1 1 0 1
# 1 1 1 1 0
qi0<-glm(Freq~OccFather+OccSon,weights=wt,family=poisson(),data=FHtab)
summary(qi0)
fitmacro(qi0)
# Quasi-independence using a "dummy factor" to create the design
# vectors for the diagonal cells (page 23).
# 1 0 0 0 0
# 0 2 0 0 0
# 0 0 3 0 0
# 0 0 0 4 0
# 0 0 0 0 5
glm.qi<-glm(Freq~OccFather+OccSon+mob.qi(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.qi)
fitmacro(glm.qi)
# Quasi-independence without using the functions
# Factor labels prevent numeric comparisons, create numeric versions
# of the row and column variables
OccFather_n <- unclass(OccFather)
OccSon_n <- unclass(OccSon)
q1 <- ifelse(OccFather_n==OccSon_n & OccSon_n==1,1,0)
q2 <- ifelse(OccFather_n==OccSon_n & OccSon_n==2,1,0)
q3 <- ifelse(OccFather_n==OccSon_n & OccSon_n==3,1,0)
q4 <- ifelse(OccFather_n==OccSon_n & OccSon_n==4,1,0)
q5 <- ifelse(OccFather_n==OccSon_n & OccSon_n==5,1,0)
glm.qi2<-glm(Freq~OccFather+OccSon+q1+q2+q3+q4+q5,family=poisson(),data=FHtab)
summary(glm.qi2)
fitmacro(glm.qi2)
# Quasi-independence constrained (QPM-C, page 31)
# Single immobility parameter
# 1 0 0 0 0
# 0 1 0 0 0
# 0 0 1 0 0
# 0 0 0 1 0
# 0 0 0 0 1
glm.q0<-glm(Freq~OccFather+OccSon+mob.qi(OccFather,OccSon,constrained=TRUE),family=poisson(),data=FHtab)
# slightly different results than Hout also found in Stata: L2=2567.658, q0=0.964
summary(glm.q0)
fitmacro(glm.q0)
# Quasi-symmetry using the symmetric cross-classification (page 23)
# 0 1 2 3 4 values of variable "sym"
# 1 0 5 6 7
# 2 5 0 8 9
# 3 6 8 0 10
# 4 7 9 10 0 */
glm.qsym<-
glm(Freq~OccFather+OccSon+mob.symint(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.qsym)
fitmacro(glm.qsym)
symmetry<-glm(Freq~mob.eqmain(OccFather,OccSon)
+mob.symint(OccFather,OccSon),family=poisson(),data=FHtab)
summary(symmetry)
fitmacro(symmetry)
# Crossings parameter model (page 35)
# 0 v1 v1 v1 v1 | 0 0 v2 v2 v2 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | 0 0 v2 v2 v2 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | 0 0 0 v3 v3 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | v3 v3 v3 0 0 | 0 0 0 0 v4
# v1 0 0 0 0 | v2 v2 0 0 0 | v3 v3 v3 0 0 | v4 v4 v4 v4 0
glm.cp<-glm(Freq~OccFather+OccSon+mob.cp(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.cp)
fitmacro(glm.cp)
# Uniform association model: linear by linear association (page 58)
glm.unif<-glm(Freq~OccFather+OccSon+mob.unif(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.unif)
fitmacro(glm.unif)
# RC model 1 (unequal row and column effects, page 58)
# Fits a uniform association parameter and row and column effect
# parameters. Row and column effect parameters have the
# restriction that the first and last categories are zero.
glm.rc1<-glm(Freq~OccFather+OccSon+mob.rc1(OccFather,OccSon),family=poisson(),data=FHtab)
summary(glm.rc1)
fitmacro(glm.rc1)
# Homogeneous row and column effects model 1 (page 58)
# An equality restriction is placed on the row and column effects
glm.hrc1<-glm(Freq~OccFather+OccSon+mob.rc1(OccFather,OccSon,equal=TRUE),family=poisson(),data=FHtab)
# Results differ from those in Hout, replicated by other programs
summary(glm.hrc1)
fitmacro(glm.hrc1)
#-------------------------------------------------------------------------------
# Examples on using these models in multinomial logistic regression
#-------------------------------------------------------------------------------
# Data from the 1972-78 GSS used by Logan (1983)
library(survival)
data(logan)
# Restructure the data in 'long' format using mlogit.data
library(mlogit)
pc <- mlogit.data(logan, shape = "wide", choice = "occupation")
head(pc,10)
# pc$alt indexes choices, pc$chid indexes respondents
# pc$occupation is a boolean variable that is TRUE where pc$alt corresponds
# with the actual choice made
class(logan$occupation) #"factor"
class(pc$alt) #"character"
# pc$alt needs to be transformed into a factor for use in the some models
# for square tables which assume ordered categories
pc$alt <- factor(pc$alt,levels=c("farm", "operatives", "craftsmen", "sales", "professional"))
# A regular multinomial logit model
m1<-mlogit(occupation ~ 0 | education+race, data=pc, reflevel = "farm")
summary(m1)
# Estimate a "quasi-uniform association" loglinear model for "focc" and "occupation"
# with "education" and "race" as covariates at the respondent level
# The quasi-uniform association model contains alternative-specific effects,
# education and race are individual specific effects
m2 <- mlogit(occupation ~ mob.qi(focc,alt)+mob.unif(focc,alt) | education+race,data=pc, reflevel = "farm")
summary(m2)
# Same model using survival::clogit
# First create a design matrix for the alternatives that drops the first category
occ.X<-model.matrix(~pc$alt)[,-1]
head(occ.X,10)
# The clogit output exceeds the default 80 characters
options(width=256)
# pc$chid which indexes respondents must be used in strata()
# Individual specific effecs are modelled as interactions between the
# design matrix for pc$alt and the covariates
cl.qu<-clogit(occupation~occ.X+occ.X:education+occ.X:race+
mob.qi(focc,alt)+mob.unif(focc,alt)+strata(chid),data=pc)
summary(cl.qu)
|
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.