Data Collection and Modelling

 Data collection for this project was mainly comprised of going through official websites and searching through any available documentation. Organizations like NAMDEVCO, the CSO and the FAO have relevant data on Trinidad and Tobago's state of food security. through the collection, processing and analyzing of associated statistical data. 

The data set from the CSO showing 'monthly imports by product', list all the imports between Jan 2020 - Apr 2020. By removing all items that aren't food related, the data set can be focused on what matters to the paper. Processing the data entailed finding the 4 month price average of the food item per kg, then finding a combined weighted average for similar food items (like combining the price of different cuts of beef and classifying it under beef). this combined weighted average per kg is divided by 10 to find the combined weighted average per 100g. The significance of 100g is that it was chosen to represent the serving size for all food items as it allows items to compared easier. As an example, an item like soy sauce is usually used by the tablespoon (14.3 grams) but with this method of comparison it can be compared to an item like sunflower oil using the same weight parameters, allowing for a more accurate comparison in terms of calories and nutrient density.  

The use of DRI constraints also play a major part in making the project easier to understand. as different nutrients have different units of measurements. In table 9, it can be seen how varied and diverse the units of measurement can become, therefore by using a percentage system for measurements a reader can tell if a DRI is being achieved or not. Full DRI is achieved by obtaining 100% of a nutrient and no DRI is achieved by receiving 0% of a nutrient (which is very difficult as the range of foods in the sample diet solutions mean that there will always be a non zero percentage for every nutrient, usually over 30% minimum). due to the selection of foods and the focus on promoting good nutrition, items like fast food are excluded. The food combinations obtained also had a low sodium content which can be good for people with health complications.  

The model itself uses 'opensolver', an extension application that runs on excel and is currently being developed by people from MIT and the University of Auckland. The software is open source and allows for people to contribute to its development. The reason why 'opensolver' is used is to allow for linear programing solutions past the 200 item limit imposed by the regular simplex solver in excel. It was quick and easy to use and helpfully outlined which constraints affected which cells, which is useful if checking for complications.