86 / 89
МОНГОЛЫН ХҮН АМЫН СЭТГҮҮЛ Дугаар (367) 20, 2011
85
All types of R-squares are relatively high in this model (Table 10).
Table 10:
Measures of Fit for MNLM of type of dwelling
Log-Likelyhood Intercept Only:
-1466.5 Log-Likelyhood Full Model:
-1138.7
D(1385):
2277.4 LR(20):
655.7
Prob > LR:
0.00
McFadden's R-square:
0.22 McFadden's Adj R-square:
0.21
ML (Cox-Snell) R-square:
0.37 Cragg-Uhler(Nagelkerke) R-square:
0.43
Count R-square:
0.64 Adj Count R-square:
0.36
RESULTS OF THE STUDY
What we have found out so far is the
variables used in MNLM have significant
effect on the type of dwelling. In addition,
model statistics are all statistically significant.
Table 11 shows the estimates of discrete
change from our model for type of dwelling.
For the
totperson
variable; each additional
person in the household, the probability of
living in
Ger
increases by 0.9 per cent and
increases the probability of living in House by
1.8 per cent while decreases the probability of
living in Apartment by 2.8 per cent, holding
all other variables at their means.
Table 11:
Discrete Change in Probability for a MNLM of type of dwelling
Variable
Change
Ger
House
Apartment
totperson
∆Range
0.229
0.127
0.216
-0.344
∆1
0.018
0.009
0.018
-0.028
∆σ
0.038
0.019
0.038
-0.057
mig_percent
∆Range
0.417
0.625
-0.231
-0.394
∆1
0.003
0.004
0.0001
-0.004
∆σ
0.064
0.094
0.002
-0.095
age
∆Range
0.262
-0.148
-0.246
0.394
∆1
0.004
-0.002
-0.003
0.006
∆σ
0.054
-0.032
-0.050
0.081
sex
0→1
0.057
0.033
0.053
-0.086
educ2
0→1
0.081
-0.073
-0.049
0.122
educ3
0→1
0.173
-0.080
-0.179
0.259
educ4
0→1
0.330
-0.171
-0.324
0.495
childnum
∆Range
0.400
0.250
0.350
-0.600
∆1
0.065
0.031
0.066
-0.098
∆σ
0.095
0.046
0.097
-0.143
totrev
∆Range
0.521
-0.306
-0.474
0.781
∆1
0.00006 -0.00004 -0.00004
0.00008
∆σ
0.114
-0.090
-0.081
0.171
food
∆Range
0.219
0.081
0.248
-0.328
∆1
0.003
0.001
0.003
-0.004
∆σ
0.044
0.016
0.049
-0.066
Note:
0→1 is change from 0 to 1; ∆1 is centered change of 1 around the mean; ∆σ is centered change of 1 standard deviation around the mean;
∆Range is change from the minimum to its maximum. is the average absolute discrete change. All the variables are held at their mean.