![Show Menu](styles/mobile-menu.png)
![Page Background](./../common/page-substrates/page0085.png)
МОНГОЛЫН ХҮН АМЫН СЭТГҮҮЛ Дугаар (367) 20, 2011
84
Table 6:
MNLM Estimation Result. Dependent variable: Type of the dwelling
Independent Variable
Ger
House
Coef. (Std. Err.)
P value Coef. (Std. Err.)
P value
Totper
0.126 (0.064)
0.05
0.108 (0.052)
0.039
mig_percent
0.037 (0.004)
0.00
0.010 (0.004)
0.017
Age
-0.028 (0.007)
0.00
-0.021 (0.006)
0.00
Sex
0.430 (0.222)
0.053
0.319 (0.172)
0.064
educ2
-0.787 (0.326)
0.016
-0.395 (0.277)
0.154
educ3
-1.186 (0.352)
0.001
-1.023 (0.295)
0.001
educ4
-2.668 (0.386)
0.00
-2.043 (0.293)
0.00
Childnum
0.437 (0.090)
0.00
0.384 (0.073)
0.00
Totrev
0.000 (0.000)
0.00 -0.0003 (0.0001)
0.00
Food
0.016 (0.008)
0.037
0.017 (0.006)
0.007
Intercept
0.271 (0.699)
0.698
0.515 (0.547)
0.347
Pseudo R-square
0.2235
LR chi-square (20)
655.67
Prob > Chi-square
0.0000
Observations
1407
From the Table 6, it is easy to see that all Z and
LR statistics are statistically significant except
constant terms
and
educ2
in case of dwelling.
For testing for independent variables, Wald test
is tighter than LR test.
Table 7 illustrates that all the coefficients
associated with the independent variables
(except
totperson
and
sex)
are statistically
significant at 0.05 level.
Table 7:
Wald test result
I n d e p e n d e n t
Variables
Chi-square
df P>Chi-square
totper
5.2
2
0.074
mig_percent
115.5
2
0.000
age
20.1
2
0.000
sex
4.8
2
0.089
educ2
5.8
2
0.05
educ3
14.8
2
0.001
educ4
65.4
2
0.000
childnum
32.4
2
0.000
totrev
45.9
2
0.000
food
7.9
2
0.02
Ho:
All coefficients associated with given variable(s) are 0.
We can reject the hypothesis that
Ger
and
House,
Ger
and Apartment, House and
Apartment are indistinguishable (Table 8).
Table 8:
Wald tests for combining alternatives
Alternatives tested
Chi-square
df
P>chi-
square
Ger
- House
88.9 10
0.000
Ger
- Apartment
259.2 10
0.000
House - Apartment
254.8 10
0.000
Ho:
All coefficients except intercepts associated with a given
pair of alternatives are 0 (i.e., alternatives can be combined).
The important test for MNLM is independence
of irrelevant alternatives (IIA). Here, I applied
Small and Hsiao test of IIA (Table 9).
Table 9:
Small-Hsiao tests of IIA assumption
Omitted lnL(full)
lnL(omit)
Chi-
square df
P>chi-
square
Evidence
Ger
-301.849 -298.031
7.635 11 0.746
for Ho
House
-174.663 -170.936
7.455 11 0.761
for Ho
Ho:
Odds (Outcome-J vs Outcome-K) are independent of other alternatives.
In our analysis, each test indicates that IIA
assumption has not been violated.
We have tested the MNLM by many tests and
the results were all statistically significant. Thus
we can apply the estimated model hereafter.