Reference no: EM132363202

Project: Computing World Happiness Index

Overview

For the last few years, the United Nations Sustainable Development Solutions Network has been publishing the World Happiness Report. Details of the 2018 report can be found here. The underlying data, which you can also download from the latter URL, is a combination of data from specially commissioned surveys undertaken by the Gallup organisation, and statistical and economic data from other sources. The web site linked above also provides the methodology for how the different data have been combined to compute the final score, most dramatically called the Life Ladder.

Here is a sample:

country	Life Ladder	Log GDP per capita	Social support	Healthy life expectancy at birth	Freedom to make life choices	Generosity	Confidence in national government
Afghanistan	2.66171813	7.460143566	0.490880072	52.33952713	0.427010864	-0.106340349	0.261178523
Albania	4.639548302	9.373718262	0.637698293	69.05165863	0.74961102	-0.035140377	0.457737535
Algeria	5.248912334	9.540244102	0.806753874	65.69918823	0.436670482	-0.194670126
Argentina	6.039330006	9.843519211	0.906699121	67.53870392	0.831966162	-0.186299905	0.305430293
Armenia	4.287736416	9.034710884	0.697924912	65.12568665	0.613697052	-0.132166177	0.246900991
Australia	7.25703764	10.71182728	0.949957848	72.78334045	0.910550177	0.301693261	0.45340696

The data shown above (and discussed below) can be found in the CSV formated text file WHR2018Chapter2_reduced_sample.csv

The actual method used to compute the Life Ladder score is quite complicated, so the the aim of this Project, in brief, is to test whether simpler methods can yield similar results. In particular, the Project aims to see whether any of a range of proposed methods yields a similar ranking, when countries are ranked by Life Ladder score in descending order i.e. from happiest on these measures, to least happy. (The Wikipedia article also discusses criticisms of the World Happiness Report process.)

Looking at the data sample above, you can see that the column headers occupy the first row, the countries are listed in the first column, while the Life Ladder scores that we are seeking to emulate are in the second column. The third and subsequent columns contain the data from which you will compute your own Life Ladder scores. However, for this exercise, please remember that the aim is not to replicate the precise Life Ladder scores, but rather to replicate the ranking of countries as a result of the Life Ladder scores.

Eye-balling the Data
In Data Science projects, it is always a good idea to "eyeball" the data before you attempt to analyse it. The aim is to spot any trends ("this looks interesting") or any issues. So, looking at the sample above (ignoring the first two columns), what do you notice?
• There is a difference in scale across the columns. Healthy Life Expectancy at Birth ranges from 52.3 to 72.8, but in general is valued in 10's, while Social Support is a value in the range 0.0 to 1.0, and Freedom to Make Life Choices has both negative and positive floating point numbers. (The problem of GDP per Capita being actually valued in the thousands, or tens of thousands, has already been solved by the data collectors taking logs.) The issue is that you don't want a particular attribute to appear significant just because it has much larger values than other attributes.
• The other thing you may have noticed is that sometimes the data is simply missing, e.g. the score for Confidence in National Government for Algeria. Any metric we propose will have to deal with such missing data (which is actually a very common problem).
• Specification: What your program will need to do

Input
Your program needs to call the Python function input three times to:
• get the name of the input data file
• get the name of the metric to be computed across the normalised data for each country. The allowed names are "min", "mean", "median" and "harmonic_mean".
• get the name of the action to be performed. The two options here are: "list", list the countries in descending order of the computed metric, or "correlation", use Spearman's rank correlation coefficient to compute the correlation between ranks according to the computed metric and the ranks according to the Life Ladder score.
The order of the 3 calls is clearly important.

Output

The output, printed to standard output, will be either a listing of the countries in descending order based on the computed metric, or a statement containing the correlation value (a number between -1.0 and 1.0).

Tasks: A more detailed specification

• Use input to read in 3 strings, representing the input file name, the metric to be applied to the data from the file (excluding the first two columns) and the action to be taken to report to the user.

• Read in the CSV formated text file. That is, fields in each row are separated by commas, e.g.
Albania,4.639548302,9.373718262,0.637698293,69.05165863,0.74961102,-0.035140377,0.457737535
Algeria,5.248912334,9.540244102,0.806753874,65.69918823,0.436670482,-0.194670126,

Apart from the first field, all the other fields are either numbers (so converted using float(), or empty, which can be translated to the Python object None. Each line will be transformed into a row, represented as a list, so you end up with a list of lists.

• For each column apart from the first two, compute the largest and smallest values in the column (ignoring any None values).

• Given the maximum and minimum values for each column, normalise all the values in the respective columns. That is, each value should be normalised by transforming it to a value between 0.0 and 1.0, where 0.0 corresponds to the smallest value, and 1.0 to the largest, with other values falling somewhere between 0.0 and 1.0. For example, the minimum Life Expectancy years in the small dataset is 52.33952713. This is transformed to 0.0. The maximum value is 72.78334045, which is transformed to 1.0. So, working proportionally, 69.05165863 is transformed to 0.81746645. In general, the transformation is (score - min)/(max-min), where max and min are the respective maximum and minimum scores for a given column, and will, of course, differ from column to column.

• For each row, across all the columns except the first two, compute the nominated metric using the normalised values (excluding None). "min", "mean" and "median" are, respectively, the minimum value (on the basis that a nation's happiness is bounded by the thing the citizens are grumpiest about), "mean" and "median" are the arithmetic mean and median value (discussed here). The harmonic mean of a list of numbers is defined here. For harmonic mean, apart from avoiding None values, you will also have to avoid any zeroes; the other metrics have no problem with 0. The output from this stage is a list of country,score pairs.

• The list of country,score pairs are either to be listed in order of descending score, or the Spearman's rank correlation coefficient should be computed between the country,score list that you have computed and the Life Ladder list, when sorted by descending score. You can assume there are no tied ranks, which means that the simpler form of the Spearman calculation can be used. An example of how to compute Spearman's rank correlation can be found here.

Important

You will have noticed that you have not been asked to write specific functions. That has been left to you. However, it is important that your program defines the top-level function main(). The idea is that within main() the program calls the other functions, as described above. (Of course, these may call further functions.) The reason this is important is that when I test your program, my testing program will call your main() function. So, if you fail to define main(), my program will not be able to test your program.

Assumptions

Your program can assume a number of things:

• Anything is that meant to be a string (i.e. a name) will be a string, and anything that is meant to be a number (i.e. a score for a country) will be a number.

• The order of columns in each row will follow the order of the headings, though data in particular columns may be missing in some rows.

What being said, there are number of error conditions that your program should explicitly test for and respond to. One example is detecting whether the named input file exists; for example, the user may have mistyped the name. The way this test can be done is to first:

import os

Then, assuming the file name is in variable input_filename, use the test:
if not os.path.isfile(input_filename) :
return(None)
and test for None in the calling function (likely main()).

Things to avoid
There are a couple things for your program to avoid.
• Please do not import any Python module, other than os. While use of the many of these modules, e.g. csv or scipy is a perfectly sensible thing to do in a production setting, it takes away much of the point of different aspects of the project, which is about getting practice opening text files, processing text file data, and use of basic Python structures, in this case lists.
• Please do not assume that the input file names will end in .csv. File name suffixes such as .csv and .txt are not mandatory in systems other than Microsoft Windows.
• Please make sure your program has only 3 calls to the input() function. More than 3 will cause your program to hang, waiting for input that my automated testing system will not provide. In fact, what will happen is that the marking program detects the multiple calls, and will not test your code at all.

Attachment:- Project and data.rar