July 9 - TRINH Thi Huong (Lucie)'s PhD defense

July 09, 2018 Campus

Thi Huong TRINH will defend her thesis in Applied mathematics on "Adapting recent statistical techniques to the study of nutrition in Vietnam" on Monday, July 9th, 2018 at 10:00 am, Room MF 323 (Manufacture des Tabacs).

SupervisorsChristine THOMAS-AGNAN and Michel SIMIONI

Memberships are:

  • Dominique HAUGHTON,  Professor and President of the jury (Bentley university)
  • Germà COENDERS,  Professor (University of Girona)
  • Phuc HO DANG,  Associate Professor (Vietnam Academy of Science and Technology)
  • Tien Zung NGUYEN,  Professor (Institut de mathématiques de Toulouse)
  • Christine THOMAS–AGNAN,  Professor and ressearcher at Toulouse School of Economics
  • Michel SIMIONI,  Senior researcher (INRA Montpellier)

The rapporteurs:

  • Dominique HAUGHTON and Phuc HO DANG


The objective of this thesis is to adapt recent statistical techniques and to bring new insights on the nutritional transition in Vietnam. Vietnam has experienced a strong economic development that turned this poor country in the 1980s into a lower middle income country currently. But Vietnam now faces the double burden of malnutrition characterized by the coexistence of undernutrition along with overweight and obesity, or diet-related noncommunicable diseases. To fight against malnutrition, the Vietnamese government has recently defined a comprehensive strategy to improve the nutritional status of the Vietnamese population.

Chapter 1 gives a brief introduction to this thesis. We consider Vietnam is a pilot case study about nutrition. We recall the main statistical techniques applied in this thesis and we emphasize our contributions.

In chapter 2, we revisit the issue of estimating the relationship between per capita calorie intake and income using six waves of the Vietnam Household Living Standard Survey over the period 2004-2014. Characterizing the response of calorie intake to income for the poorest households is a prerequisite for considering policies aimed at reducing starvation and correcting nutritional deficiencies. The classical log-log specification does not capture the nonlinearity of this relationship. To avoid the curse of dimensionality of fully nonparametric specifications due to the presence of many control variables (age, education, region …) we adopt rather various generalized additive models (GAM) specifications where only income is supposed to act in a nonlinear fashion and compare them with a recent procedure. The results highlight the strong response of calorie intake to an increase in income for the poorest households. A byproduct of the proposed methodology is the decomposition of the evolution of average calorie intake between the two waves into the part due to the change of population characteristics distributions and those coming from the change in calorie-income relationship, shedding new light on the nutritional transition in Vietnam.

In Chapter 3, we use decomposition methods to assess the determinants of changes in macronutrients consumption in Vietnam using the 2004 and 2014 waves of VHLSS. The common objective of decomposition methods is to decompose between-group differences in economic outcomes such as wage or income, into two components: a composition effect due to differences in observable covariates across groups, and a structure effect due to differences in the relationship that links the covariates to the considered outcome. The recent decomposition procedure proposed by Rothe (2015), which can be applied to mean, quantiles, or other parameters characterizing the distribution of the considered outcome, aims at decomposing further the composition effect into three types of components: (1) the direct contribution of each covariate due to between-group differences in their respective marginal distributions, (2) several two way and higher order interaction effects due to the interplay between two or more covariates and (3) a dependence effect accounting for different dependence patterns among the covariates. Rothe (2015) uses a parametric copula to model the dependence effects, which is well adapted for continuous covariates. We adapt this approach to the case of a mixture of continuous and discrete covariates.

In Chapter 4, we focus on food composition in terms of diet components. We consider modeling the proportions of protein, fat and carbohydrate (D=3) in the average per capita calorie intake. Because this vector of proportions is of a compositional nature, we naturally turn attention the compositional data analysis techniques. We use descriptive tools, such as compositional biplots and ternary diagrams, to show the evolution of the three components over the years and then model macronutrients composition as a function of household characteristics, using compositional regression models. We derive the expression of the semi-elasticities of macronutrients shares with respect to food expenditure. We then compare the interpretations of these shares semi-elasticities to that of volumes of macronutrients and of total calorie intake obtained using classical linear models.

In Chapter 5, we focus on the relationship between macronutrient balances and body mass index. We develop a compositional regression model including a total at various quantile orders. This approach solves the problem of confounding effects between macronutrients total volume and shares in a diet (Willett et al., 1997). We then compute the elasticities of BMI with respect to each macronutrient and to the total consumption. Our empirical research is based on the General Nutrition Survey 2009-2010 and we restrict attention to Vietnamese adults from 18 to 60 years of age. The results first reveal significant impacts of some socio--economics factors, such as the overall consumption volume, the age, the gender, the job type, the ``no drinking status'' and the geographical region. All elasticities of BMI with respect to each macronutrient increase as BMI increases until a threshold (BMI=20) and then remain stable.

In chapter 6, we briefly give our perspectives of future research in both mathematics and nutrition.