69 the nature of logistic function. In other word, the altitude and slope might not
become people’s consideration to transform the existing land use into other land uses to fulfill their needs in Siak District. The similar consideration might also
happen for distance from health service and the area of sub district which both variables did not contribute significantly to the land use change in Siak District.
For instance, the area of sub district and distance from health service were not considered by people in transforming the forest to crop plantation or from
agriculture area to settlement area. On the contrary, the development of settlement may be able to motivate government to develop the health service near the
settlement area. Table 15. Likelihood Ratio Tests for All Significant Variables
Likelihood Ratio Tests
29535.707 158.726
25 .000
29454.986 78.005
25 .000
31247.253 1870.272
25 .000
30628.102 1251.121
25 .000
31781.668 2404.687
25 .000
29578.108 201.127
25 .000
31779.854 2402.873
25 .000
29431.426 54.445
25 .001
29488.730 111.749
25 .000
29427.464 50.483
25 .002
29429.034 52.053
25 .001
29418.559 41.578
25 .020
29441.404 64.423
25 .000
29475.306 98.325
25 .000
29417.749 40.768
25 .024
29425.812 48.831
25 .003
29442.620 65.639
25 .000
29473.133 96.151
25 .000
29433.467 56.485
25 .000
29510.285 133.304
25 .000
29570.020 193.039
25 .000
29451.609 74.628
25 .000
29467.936 90.955
25 .000
29449.923 72.942
25 .000
29460.043 83.062
25 .000
Effect Intercept
RIVDIST FRSTDIST02
CROPDIST02 GRASDIST02
WETDIST02 OTHEDIST02
KWSID RTRWPID
CONCESSID POPDENS02
ANIMDIST ECONDIST
EDUCDIST ENVIDIST
FISHDIST GOVDIST
INDUSTDIST MININGDIST
ROADDIST SETTLEDIST
TELCOMDIST TRANSMDIST
TRANSPDIST PUBSPADIST
-2 Log Likelihood of
Reduced Model
Model Fitting Criteria
Chi-Square df
Sig. Likelihood Ratio Tests
The chi-square statistic is the difference in -2 log-likelihoods between the final model and a reduced model. The reduced model is formed
by omitting an effect from the final model. The null hypothesis is that all parameters of that effect are 0.
70
1.4.2.2 Significance Test for Final Model 1
st
Scenario
In the previous activity, the MLR model analysis has determined which variables could contribute significantly into the land use change model. From 28
variables that have been analyzed, there are 24 variables considered as important variables which have driven the land use change in Siak District during 2002 –
2005. The MLR model analysis has also simulated the final model of land use change in Siak District by considering the important variables found. There were
two tests which have been done in order to determine whether the final model were adequate to explain the land use change happen in Siak District or not. The
tests done were Likelihood Ratio Test of the final model and Pseudo R-Square statistics.
The model fitting information in SPSS outputs shows the Likelihood Ratio Test of the Final model against one in which all the parameter coefficients are 0
Null. A likelihood ratio test shows that the model fits the data better than a null model. Based on the model fitting information which has been produced, the
significance level of the likelihood ratio test for the final model of land use change in Siak District is 0.000. Its significance level is less than 0.05 Sig.0.05 which
may be concluded that the final model which has been produced was outperforming the Null. In this case, by considering the significant variables
driving factors determined before, the land use change model of Siak District that has been developed might fit the data better than a null model, and it could be
concluded as a good fit model.
Table 16. Likelihood Ratio Test of the Final Model 1
st
Scenario
Model Fitting Information
53960.379 29376.981
24583.398 600
.000 Model
Intercept Only Final
-2 Log Likelihood
Model Fitting
Criteria Chi-Square
df Sig.
Likelihood Ratio Tests
71 Table 17. Pseudo R-Square Statistics 1
st
Scenario
Pseudo R-Square
.914 .919
.456 Cox and Snell
Nagelkerke McFadden
The pseudo r-squared statistics are designed to have similar properties to the true r-squared statistic, which measures the variability in the dependent
variable that is explained by a linear regression model. Larger pseudo r-squared statistics indicate that more of the variation is explained by the model, to a
maximum of 1. There were three approaches of pseudo r-squared statistics which have been done in this research which have been proposed by Cox and Snell,
Nagelkerke, and McFadden. Based on the computation, pseudo r-squared statistics, either Cox and Snell or Nagelkerke, indicate that the variability of the
land use change in Siak District as dependent variable can be explained more than 0.91, whereas according to McFadden, it is only 0.456. In other word,
according to Cox and Snell and Nagelkerke, more than 91 of the variability of land use change in Siak District can be explained by the final model developed,
and McFadden indicate only 45.6 of the variability can be explained by the model. However, in general the pseudo r-squared statistics that have been
produced indicates that most of variability in land use change which happen in Siak District could be explained by the final model.
The two tests for final model which have been conducted, likelihood ratio test for the final model and pseudo r-squared, indicate that the final model of land
use change in Siak District which has been developed by using MLR model is a good model that can explain most of the variability of land use change in the
research site. The significant variables that have been determined can be concluded as general driving factors which drive the land use transitions in Siak
District. Furthermore, these research findings, which have been interpreted from the statistical computation, should be spatially compared to the actual condition in
order to validate the performance of the MLR model when the model is applied in spatial manner.