FIVIMS: Food Insecurity and Vulnerability Information and Mapping Systems FSAU: Food Security Analysis Unit (FAO – Somalia). Food security experts and international organizations have considered Food Consumption Scores as a proxy measure of food insecurity.
Introduction
The United Nations World Food Program (WFP) described South Sudan as “the poorest and least developed region in Sudan and one of the poorest and least developed regions in the world” (WFP-VAM, 2007b). Although the word 'health' was central to the study design (UNICEF mainly funded the activity), a separate questionnaire on food security was administered, but the data was not formally analyzed and the report did not address food security information (GoSS ). -MOH, 2008).
Importance of the study
Measuring food insecurity at the household level has preoccupied even food security analysts for years. Another value of the analysis could be to provide baseline information for subsequent future assessments of the food security situation.
Research problem statement
Further, the study attempts to identify appropriate options to intervene in food security related issues in South Sudan. Furthermore, the study is also expected to guide the development of policies and strategies to improve food security in South Sudan.
Sub-problems
Finally, the study aims to provide a basis for further research on those variables determined to be statistically significant in their association with food insecurity outcomes.
Hypotheses
Conceptual framework for describing the rationale of the study
Study limitations
Definition of terms
A condition that exists when all people, at all times, have physical, social and economic access to sufficient, safe and nutritious food that meets their dietary needs and food preferences for an active life" (FAO, 1996). .
Study assumptions
Methodology
Dissemination of Results
Organisational structure of the mini-dissertation
Summary
Introduction
Review of existing approaches in measurement of food insecurity
The search for a “Gold Standard” for measuring household food insecurity
Assess poor people's perceptions of food insecurity and measure the experience of hunger (Coates et al., 2003). Enables the examination of food insecurity at the household and intra-household level (Swindale and Bilinsky, 2006).
Approaches for measurement of the determinants of household food insecurity
Further, he categorizes the causes of food insecurity as: (a) availability of quantity and quality of household food; (b) physical and economic access to food. Riely, et al (1999) suggests that the causes of food insecurity can be understood in the context of a food security framework.
Introduction
The binomial distribution and the standard normal distribution
- The probability distribution of a ‘success’ response
- Methods of inference based on the ‘success’ probability
- Logistic Regression Model for Binary Data
- The response data and preliminary modelling concepts
- The Logistic transformation
The binomial distribution of the random variable Y, denoted as B(n,p), suggests that the variance of an observation y is np(1-p). Rejection of the null hypothesis is suggested if the perceived probability of success is too small.
The linear Logistic Regression model
Fitting the linear Logistic Regression model to binomial data
Since the form of the data distribution is binomial, we maximize the distribution likelihood function described in subsection 3.1.1. It should be noted that this probability is a function of (the vector of 's) as it depends on the unknown which in turn depends on the β's through equation (3.9). Fitting the data to a logistic regression model in SPSS will generate the k+1 parameter estimates described above, as well as the odds ratio estimate and the predicted probability of “success”.
The standard errors of parameter estimates
These fitted individual probabilities are later compared to the observed probabilities for each group and each category of the response variable to evaluate the validity of the fitted model. As an illustration, we examine a statistical model that describes the relationship between the probability of a household becoming food insecure and sources of livelihood namely;. Let us assume that the interest is centered on estimating the success probability of a household becoming food secure at each level of a life path.
Testing for the significance of the model
In the likelihood ratio test, the degrees of freedom are equal to the number of additional parameters in the model with a factor. Using this information, we can then determine the critical value of the test statistic from standard statistical tables. The score test is equivalent to the Pearson chi-square statistic described in Chapter 3.
The Logistic Regression Model for Ordered Categorical Data
- Formulation of the proportional odds model for ordered categorical data
- Comparison between two households
- The Mann-Whitney test of the proportional odds model
- Fitted Probabilities
- Calculating the Deviance
- Hypothesis testing
- Model Checking
However, due to the nature of the data available for this project, only the proportional odds model will be discussed here. The proportional odds model can be understood as an extension of logistic regression or, as proposed by Collet (2003), a “generalization of the logistic regression”. Further understanding of the theory on the formulation of the Proportional Odds Model (other texts use 'Cumulative Odds Ratios') can be found in Agresti (2002), McCullagh (1980), Peterson and Harrel (1990) and SPSS (2006).
Introduction
Description of the dataset
Sample selection, data collection and processing
If no suitable respondent was found, only the household questionnaire was completed by asking an adult in the household to answer the questions. In the event that the household turned out to have no one, a neighbor was asked whether it was occupied. If the household was occupied, the research team asked the neighbor when the household members would return.
Derivation of the main response variable
A series of steps leads to the calculation of the variable Food Consumption Score or Household Dietary Diversity Score. The number of times a food is eaten during a week or the frequency of food intake and the standard weight of the food group form the basis for calculating the FCS. The Food Consumption Score was then calculated for each household by summing the product of the frequency of FCG multiplied by the corresponding weight.
The set of predictor variables
Finally, the degree of food insecurity for each household is determined based on the food consumption profile, so that “poor”, “marginal” and “good” food consumption behavior can represent poor, marginal and acceptable levels of food security, respectively.
The data analysis techniques
The SPSS PLUM technique will be used to: (i) investigate model fit; (ii) generate parameter estimates for determining the difference between categories of the response variable; (iv) calculate the fitted probability values; (v) provide the inspection model of eligibility statistics and; (vi) produce hypothesis tests of the significance of relationships or associations between categories of the response variable and levels of significant predictors. The SPSS analysis will involve the application of a Logistic Regression technique called Ordinal Regression, which is suitable for the Proportional Odds Model. Interpreting the results of Ordinal Regression procedures will enable a deeper understanding of the results and findings.
Model selection
Forward Selection
For each of the independent variables, the method calculates an F-statistic that reflects the contribution of the variable to the model if it is included. The p-values for these F-statistics are compared to the SLENTRY = value specified in the model statement (or to 0.50 if the SLENTRY = option is omitted). Variables are added to the model one at a time until no remaining variable produces a significant F-statistic.
Backward Elimination
It then recomputes the F-statistics for the variables still outside the model and repeats the evaluation process.
Stepwise Selection
Of the three methods of model selection, the backward elimination method is chosen simply because of automation. Additional advantages of the backward elimination method over the other criteria are described in the linear regression literature. Similarly, it cannot be included for the backward elimination algorithm once a variable has been removed.
Procedures for model checking and diagnostics
The Score Test for validation of the proportional odds assumption
According to Hocking (2003), both the forward selection and backward elimination criteria have lent themselves to criticism. He indicates that the forward selection criterion is seen as weak because once a variable has been entered it cannot be removed. Singh (2004) notes that the limitation of both methods is that once a variable is removed, their meaning may change.
Fitted probabilities and frequencies
SPSS allows comparison of fitted probabilities with observed probabilities using a procedure known as a classification table. We can conclude that the higher the number of correctly predicted categories, the better the fit of the model. In interpreting the findings, if the observed and adjusted probabilities are similar for the predictor variables examined, this reflects good model fit.
Direct assessment of the model assumption for the proportional odds model
Introduction
For this reason, annotated analysis using the Linear Regression technique will be presented and compared (Section 5.3). As explained in Chapter 4, the backward elimination strategy will be used for the selection of the predictor variables of influence. Results of the electronic outputs of the procedures will be discussed and appropriate interpretations of useful findings will be made.
Exploratory analysis
- Exploratory analysis based on food consumption score as a continuous variable
- Exploratory analysis of linear relationships based on correlation statistics
- Exploratory analysis based on food consumption score as a discrete ordinal variable
- Conclusion
This subsection aims to inspect the distribution of food consumption score (FCS) observations from a sample size of 9 220 households. The graph (see Figure 5.2) shows that the distribution of food consumption scores tends towards normality, with a slight skew towards the bottom of the scale. Results of testing for the significance of correlations between the food consumption score and each of the three covariates vary between the two types of tests.
Logistic Regression analysis based on the Proportional Odds Model
- Choice of a Link Function
- Fitting the ordinal logistic regression to the food consumption data
- Running the analysis
- Evaluating the model
- Test of parallel lines
- Interpreting the model
- Revising the model
- Classification Table of the final model
- Results and discussion
- Conclusion
This is an indication of the improvement of the model with predictors compared to the one without predictors. In the parallel lines test, the chi-square value increases. However, this did not affect the significance of the difference in predictor levels associated with food consumption categories.
Fitting of Linear Regression model to the continuous response variable
- Important assumptions of the Linear Regression model
- Exploration of linear relationship
- Inspection of the fitness of the model
- Interpretation of the model coefficients
- Conclusion
The third and final assumption is that the error term value for each case (household or individual) is independent of the values of the model variables in the model and of the values of the error term for other cases (SPSS, 2006). Based on the two plots alone, it is necessary to get the impression that the data set is not suitable for use of the linear regression model. There is sufficient statistical evidence to suggest that the five eliminated independent variables, shown in the boxes in Table 5.15 above, do not contribute significantly to the model, regardless of whether linear regression or ordinal regression is used.
Introduction
Conclusions
- Proportional Odds Model appropriate for predicting food consumption outcomes
- At least eleven factors influenced food insecurity in Southern Sudan
- At least eight factors could be used for food insecurity surveillance
- Easily replicable methodology
- Peculiar findings
The study identified at least eleven factors as significant predictors of food consumption in more than one set of analyses; hence the food insecurity in South Sudan (at least in 2006 or immediately after the end of the 21-year civil war). The finding that food aid was not an important factor in food consumption is extremely interesting. In other words, poor food consumption and nutrition shocks—at least for the study period—characterized South Sudan.
Recommendations
Pietermaritzburg, African Center for Food Security (ACFS), University of KwaZulu-Natal (UKZN). 2002) Sustainable Livelihoods Approach: Concept and Practice, Palmerston North, Massey University. The challenges facing empirical estimates of household food (in)security in South Africa, Development Southern Africa. Washington DC, International Food and Policy Research Institute (IFPRI). 1999b) Operationalizing household food security in development projects: an introduction.
Please, after completing LAP 1-7, complete the following table 1 crop at a time (line by line) and repeat the most frequently asked questions. HOW MUCH DID YOU EAT OF THIS PRODUCT COMPARED TO THE LAST WEEK IN THE LAST HARVEST SEASON. HOW MUCH OF THIS LIFE DID YOU EAT IN THE PREVIOUS RAINY SEASON COMPARED TO THE LAST WEEK.
LAP4 It is worth finding out how households fared in terms of food consumption in relation to farm food collection. Therefore, it is probably a matter of knowing the magnitude of the impact and importance of this variable on food consumption levels. FAI1 It is worth investigating whether receiving food aid had a significant effect on FCS.
METHOD 3: The linear regression technique for fitting a model where the dependent variable is a ratio scale (continuous). SOMELY EDITED SPSS ORDINAL REGRESSION OUTPUT FOR A MODEL FITTED WITH COMPLEMENTARY LOG-LOG LINK FUNCTION. There are cells (ie, dependent variable levels at combinations of predictor variable values) with zero frequencies.
