Versión de HTML Básico
Lorena Moreno
96
Analiti a, Revista de análisis estadístico, Vol. 13 (1), 2017
Table 2:
Stunting and HAZ average by BDH reception
Group
Observations Stunting Prevalence HAZ
Non-beneficiaries
4.197
22,90
-1,11
(-0,006)
(-0,020)
Beneficiaries
1.977
39,20
-1,61
(-0,011)
(-0,028)
N
6.174
t
-13,487
14,560
Degrees of freedom 6172
p-value
0,000
0,000
Note:
t statistics for the null hypothesis of no significant differences between
group means (beneficiaries and non-beneficiaries) in stunting prevalence and
HAZ have associated p-values
<
0,001. Therefore, we can reject the null.
Alternating Least Squares was applied, because it allows the choice of both, measurement
level and a number of sets. The combination of categorical optimal scaling level and one
set of variables was chosen by the programme designing team. This specification lead to a
Categorical Principal Components Analysis (CAPTCA) procedure, which was implemented
by the SPSS algorithm named the same way (Fabara, 2009).
In Ecuador, the indexes were built based on variables highly correlated with the per-capita
monthly consumption aggregate and optimally quantified in 2 dimensions. The matrix of
the subjects in rows and these variables in the columns is the initial input for the calculation.
The procedure itself consists of two iterative phases; first, parameter estimation and second,
optimal scaling. In this way, scores with maximum heterogeneity between categories and
maximum within-category homogeneity are assigned to subjects (Guerrero, 2002). Once
the parameters (or categorical quantifications) are estimated, the category with the lowest
quantification is given a value of 0 and the remaining quantifications are subtracted this
original lowest weight. Finally, these new values are rescaled from 0 to 100 (SIISE, 2014).
The method was the same throughout the years, though the choice of variables changed
towards a more structural and less income based measure of deprivation. The first index,
SELBEN, was estimated with 27 variables with the 1999 LSMS data and was valid until 2007.
Afterwards, in 2008, the RS index was calculated with the 2006 LSMS including 30 items. As
part of the two Social Registry information gathering processes, a survey was implemented in
early stages, called the Households Socio-Economic Situation Survey (HSSS). Once the first
RS official database was collected in 2009, the 2006 LSMS index was updated on the basis
of the first HSSS. For BDH’s third stage, the RSII was calculated directly with the second
HSSS, collected in 2012. The 2012 HSSS has statistical representativeness at national, urban
14