For this specific objective your dependent variable will be Food Security Scores/Index and independent variable will be consumption expenditure patterns. If you could explain how is the food security measured and what are the range of values, it would be possible to suggest some models otherwise it is difficult to do so. If you use a dummy for food security with values zero and one, you can use regression models with dummy dependent variables. If you use a multiple choice questions for the dependent variables, then multinomial logit type model will be advisable. However, the method of estimation will also differ depending on whether your model has problems of Endogeniety, collinearity, etc.

hope the above will help you.

The model to use strongly depends on how the food security of the household is measured. A logitic regression model can be applied if the food security of the household (the dependent variable) is measured as dummy variable. However, the the food security of the household is measured as a continuous variable (level of intake of food compared the the standard intake requirement) an OLS model may be applied. You may take a look at this paper. Hope it will be of benefit to your work.Article Analysis of the Determinants of Calorie Intake Among Rural H...

Najibullah Hassanzoy thank you for your response, which is very helpful.

Regarding your query, I measured food security by undertaking a 7-day recall of food consumption for the individuals within a household (that is for both french beans producing and non-producing households). To determine if a household was food secure or not, I compared the calculated per capita calorie intake with the kilocalories threshold.

Oluwafemi Enilolobo, thank you for your advise which is so helpful and for sharing with me the paper. I will read it with keen interest.

Salamona Hassanzoy saib, hope you are well. I agree with Lydiha here, just would like to add that the best technique of all times is OLS! unless it is LMD problem (i.e.the dependent variable is censored or is binary or is a count variable for which OLS is not a suitable fit as it ignores censored or binary properties of the data). Par capita caloric intake seems like a continuous variable and for that OLS is the best fit, unless some observations consume zero calories associated french beans or something, then you need to use a tobit or other flexible forms of tobit such as craggit (if need be).

Dear Ahmadzai,

Note that in your context, you are expressing food security as a function of household expenditure pattern and other relevant variables. Remember that household expenditure pattern is influenced by a set of variables which may create a potential endogeneity issue. For this reason, it is important to consider an instrumental variable approach. As indicated by previously by some honourable colleagues, it is important to rethink how you want to operationalise food security. The operationalisation depends on the dimension of food security you want to analyse. You can email me for further deliberation and assistance.

Best regards

Emmanuel