Time Series Problem Set 1
Carrie Coberly
Feb. 3, 2017
Introduction
This problem set prepares the data for, conducts longitudinal and cross-sectional comparisons of, and performs basic data visualization comparing regime type and basic health statistics across 161 countries for the years 1960-2015. The HTML file for this assignment can be found at: http://rpubs.com/carriecoberly/247473
Part 1: Data Management
Loading the data and the relevant libraries
First, I load the necessary libraries in R:
library(dplyr)
library(tidyr)
library(ggplot2)
suppressPackageStartupMessages(library(googleVis))
library(plotly)
library(knitr)I am using the Varieties of Democracy dataset for information on regime type, and data from the UN and World Health Organization on the number of physicians, percent malnourished, and health expenditures across countries. Both datasets are comma separated values files. I load both files:
vdem <- read.csv("vdem.csv")
health <- read.csv("UNdata.csv")Cleaning the data
Varieties of Democracy
For this assignment, I will examine the relationship between health indicators and two measures of democracy, three measures of civil liberties, and three measures of women’s rights. To do so, I will keep the following variables from the Varieties of Democracy dataset:
- Country Identifiers
- country_name
- country_text_id: The UN country identifier
- COWcode: The country code used by the Correlates of War database
- Measures of Regime Type
- v2x_polyarchy: This measure of electoral democracy (or polyarchy) evaluates whether a country hold free and fair elections. It incororates measures of suffrage, freedoms of association and speech, and electoral fraud. This variable is an interval variable - a continuous measures of democracy scaled between 0 and 1.
vs2_libdem: This measure of liberal democracy incorporates the definition of electoral democracy with additional protections of individual and minority rights. In addition to the freedoms described above, this variable also examines whether there are constitutionally protected civil liberties, rule of law, an independent judiciary, and effective checks and balances on the executive. This variable is an interval variable - a continuous measures of democracy scaled between 0 and 1.
- Measures of Civil Liberties
- v2x_cspart: This index evaluates whether major civil society organizations are routinely consulted by policymakers; popular participation in civil society organizations, including female participation; and the centralization of candidate nominations within parties. This variable is an interval variable - a continuous measures of participation scaled between 0 and 1.
- v2xme_altinf: This index evaluates the extent to which print and broadcast media are un-biased in their coverage, allowed to be critical of the regime, and representative of a wide array of political perspectives. This variable is an interval variable - a continuous measure of bias scaled between 0 and 1.
v2xcl_rol: This index measures the extent to which laws are transparent, rigorously and impartially enforced and the extent to which citizens enjoy access to justice and basic rights. It is formed using idicators for impartial public administration, transparent and predictable laws, access to justice, property rights, freedom from torture, political killings and forced labor, and freedom of religion and movement. This variable is an interval variable - a continuous measure of rule of law scaled between 0 and 1.
- Measures of Gender Equality
- v2xeg_eqprotec: This index measures how equal the protection of rights and freedoms is across social groups. This variable is a ratio variable - a continuous measure of protection scaled between 0 and 1.
- v2x_gender: This variable measures women’s political empowerment, which incorporates protection of women’s civil liberties, women’s and participation in civil society organizations, and the representation of women in the formal political system. This variable is an interval variable - a continuous measure of rights scaled between 0 and 1.
v2x_suffr: This variable measures the percent of adult citizens that has the legal right to vote in national elections. It is an interval variable - a percentage between 0 and 1.
vdem <- select(vdem, country_name, country_text_id, COWcode, year, v2x_polyarchy, v2x_libdem, v2x_cspart, v2xeg_eqprotec, v2x_gender, v2xme_altinf, v2x_suffr, v2xcl_rol)The Varieties of Democracy includes information on territories not recognized by the UN (or included in their datasets). In order to remove these territories (which do not have Correlates of War codes) from the dataset, I use the code:
vdem <- drop_na(vdem, COWcode)This dataset is now ready to merge with the UN data.
UN Health Data
The UN data is organized with years as variables instead of observations. In order to format this data to merge with the Varieties of Democracy data, I need to transpose the columns and rows and reformat them to match the VDEM data.
First, I will remove one unnecessary variable:
health <- select(health, -Series.Code)The column that includes variables names also includes rows that do not include relevant information. To eliminate them, I will filter to keep only rows that have data associated with our three variables of interest: Health expenditure per capita, Physicians per 1,000 people, and Prevalence of undernourishment. I also remove data for UN aggregate measures and territories.
levels(health$Series.Name)## [1] ""
## [2] "Data from database: Health Nutrition and Population Statistics"
## [3] "Health expenditure per capita (current US$)"
## [4] "Last Updated: 12/16/2016"
## [5] "Physicians (per 1,000 people)"
## [6] "Prevalence of undernourishment (% of population)"health <- filter(health, Series.Name=="Health expenditure per capita (current US$)" | Series.Name=="Physicians (per 1,000 people)" | Series.Name=="Prevalence of undernourishment (% of population)")
health <- filter(health, Country.Code!="ARB" & Country.Code!="ASM" & Country.Code!="CSS" & Country.Code!= "CEB" & Country.Code!= "EAR" & Country.Code!="EAS" & Country.Code!="EAP" & Country.Code!="TEA" & Country.Code!="EMU" & Country.Code!="ECS" & Country.Code!="ECA" & Country.Code!="TEC" & Country.Code!="EUU" & Country.Code!="FRO" & Country.Code!="FCS" & Country.Code!="PYF" & Country.Code!="GIB" & Country.Code!="GRL" & Country.Code!="GUM" & Country.Code!="HPC" & Country.Code!="HIC" & Country.Code!="HKG" & Country.Code!="IMY" & Country.Code!="LTE" & Country.Code!="LAC" & Country.Code!="LCN" & Country.Code!="LDC" & Country.Code!="LMY" & Country.Code!="LIC" & Country.Code!="LMC" & Country.Code!="MAC" & Country.Code!="MEA" & Country.Code!="MNA" & Country.Code!="TMN" & Country.Code!="MIC" & Country.Code!="NCL" & Country.Code!="NAC" & Country.Code!="MNP" & Country.Code!="OED" & Country.Code!="OSS" & Country.Code!="PSS" & Country.Code!="PST" & Country.Code!="PRE" & Country.Code!="PRI" & Country.Code!="SXM" & Country.Code!="SST" & Country.Code!="SSA" & Country.Code!="SSF" & Country.Code!="SAS" & Country.Code!="TSA" & Country.Code!="TSS" & Country.Code!="MAF" & Country.Code!="WLD" & Country.Code!="UMC" & Country.Code!="VIR" & Country.Code!="VGB" & Country.Code!="WBG" & Country.Code!="CYM" & Country.Code!="CHI" & Country.Code!="TLA")We need to reshape the data into a long version by moving the years to the rows. We use the gather command for that:
health <- gather(health, key="year", value="stat", X1960..YR1960.:X2015..YR2015.)## Warning: attributes are not identical across measure variables; they will
## be droppedThe UN data categorized missing data with a “..” instead of the “NA” standard for R. To convert the data, I place the data into a separate vector, replace all values as appropriate within that vector, append that vector to the original dataset and drop the earlier version of the variable. I then can convert this column to a numeric variable.
conv <- health$stat
replace(conv, conv=="..", NA)
health$conv <- conv
health <- select(health, -stat)
health$conv <- as.numeric(health$conv)The year variable still has an X prefix and a suffix. To remove this prefix we treat year like a character variable, and use the substr command to omit the unneeded characters. Finally we convert year to a numeric variable.
health$year <- substr(health$year, 2, 5)
health$year <- as.numeric(health$year)I now need to move the three variables to columns:
health <- spread(health, key=Series.Name, value=conv)For clarity and to match the variable names to the VDEM data, I rename the variables as follows:
names(health) <- c("country", "country_text_id", "year", "health.exp", "physicians", "undernour")Since the “undernour” variable is the percentage of the population that is undernourished, I will convert it to a fraction between 0 and 1 using the mutate command:
health <- mutate(health, undernour = undernour/100)Merging the datasets
To merge the two datasets, I use the “join” command. The “full_join” command keeps all data, regardless of whether they match or not.
vdem_un <- full_join(vdem, health, by=c("country_text_id", "year"))## Warning in full_join_impl(x, y, by$x, by$y, suffix$x, suffix$y): joining
## factors with different levels, coercing to character vectorThe UN health data had information on more countries than are included in the VDEM data (countries with populations smaller than 1 million people, primarily Pacific and Caribbean island countries), and includes data on countries prior to their independence (for example, on Armenia before the breakup of the Soviet Union). The VDEM data includes historical countries (such as South Yemen and North Vietnam) that are not in the UN health data. I would normally go back to the codebook for the UN data to determine whether the coders separated data for the Soviet Union into each republic to see if countries need to be reconstructed, but in this case I do not have that codebook and choose to exclude data on countries that do not exist in the other dataset. This will eliminate small countries, historical countries, and countries prior to independence from the dataset.
vdem_un <- drop_na(vdem_un, country)
vdem_un <- drop_na(vdem_un, country_name)The last step is to rearrange the columns and rows. I will do this by dropping the extra country name variable, after this the data will be in order. I use arrange() to sort the data by country name and year.
vdem_un <- select(vdem_un, -country)
vdem_un <- arrange(vdem_un, country_name, year)In order to find out how many countries remain in the dataset, and how many years are covered for each country, I use the “count” command:
count(vdem_un, country_name)## # A tibble: 161 × 2
## country_name n
## <fctr> <int>
## 1 Afghanistan 56
## 2 Albania 53
## 3 Algeria 56
## 4 Angola 53
## 5 Argentina 56
## 6 Armenia 26
## 7 Australia 55
## 8 Austria 53
## 9 Azerbaijan 26
## 10 Bangladesh 44
## # ... with 151 more rowsThe dataset now has data for 161 countries for the years 1960-2015, although many countries are missing data for one or more years.
Part 2: Collapsing the data
I will now examine whether the percent undernourished in a country (undernour) is related to the country’s overall level of electoral democracy (v2x_polyarchy) and women’s empowerment (v2x_gender).
I will first examine the data by year. To do so, I group (group_by()) and provide summary statistics of (summarize()) the data by year.
vdem_un_year <- group_by(vdem_un, year)
vdem_un_year <- summarize(vdem_un_year, demmean=mean(v2x_polyarchy, na.rm=TRUE),
gendermean=mean(v2x_gender, na.rm=TRUE),
nourmean=mean(undernour, na.rm=TRUE))
kable(vdem_un_year, digits=3)| year | demmean | gendermean | nourmean |
|---|---|---|---|
| 1960 | 0.329 | 0.412 | NaN |
| 1961 | 0.332 | 0.412 | NaN |
| 1962 | 0.330 | 0.410 | NaN |
| 1963 | 0.333 | 0.421 | NaN |
| 1964 | 0.330 | 0.429 | NaN |
| 1965 | 0.331 | 0.430 | NaN |
| 1966 | 0.332 | 0.437 | NaN |
| 1967 | 0.327 | 0.442 | NaN |
| 1968 | 0.325 | 0.444 | NaN |
| 1969 | 0.322 | 0.449 | NaN |
| 1970 | 0.322 | 0.467 | NaN |
| 1971 | 0.321 | 0.471 | NaN |
| 1972 | 0.321 | 0.468 | NaN |
| 1973 | 0.319 | 0.457 | NaN |
| 1974 | 0.319 | 0.470 | NaN |
| 1975 | 0.324 | 0.501 | NaN |
| 1976 | 0.328 | 0.509 | NaN |
| 1977 | 0.331 | 0.510 | NaN |
| 1978 | 0.336 | 0.511 | NaN |
| 1979 | 0.350 | 0.512 | NaN |
| 1980 | 0.356 | 0.522 | NaN |
| 1981 | 0.354 | 0.528 | NaN |
| 1982 | 0.354 | 0.527 | NaN |
| 1983 | 0.357 | 0.533 | NaN |
| 1984 | 0.362 | 0.533 | NaN |
| 1985 | 0.374 | 0.536 | NaN |
| 1986 | 0.380 | 0.545 | NaN |
| 1987 | 0.385 | 0.547 | NaN |
| 1988 | 0.393 | 0.554 | NaN |
| 1989 | 0.401 | 0.564 | NaN |
| 1990 | 0.437 | 0.598 | NaN |
| 1991 | 0.469 | 0.611 | 0.245 |
| 1992 | 0.493 | 0.617 | 0.248 |
| 1993 | 0.506 | 0.627 | 0.248 |
| 1994 | 0.514 | 0.637 | 0.248 |
| 1995 | 0.518 | 0.644 | 0.244 |
| 1996 | 0.524 | 0.645 | 0.240 |
| 1997 | 0.527 | 0.649 | 0.235 |
| 1998 | 0.528 | 0.657 | 0.232 |
| 1999 | 0.526 | 0.661 | 0.226 |
| 2000 | 0.529 | 0.677 | 0.220 |
| 2001 | 0.531 | 0.682 | 0.213 |
| 2002 | 0.541 | 0.688 | 0.208 |
| 2003 | 0.544 | 0.689 | 0.202 |
| 2004 | 0.546 | 0.695 | 0.196 |
| 2005 | 0.551 | 0.701 | 0.190 |
| 2006 | 0.555 | 0.702 | 0.183 |
| 2007 | 0.552 | 0.708 | 0.177 |
| 2008 | 0.558 | 0.712 | 0.171 |
| 2009 | 0.561 | 0.713 | 0.166 |
| 2010 | 0.561 | 0.717 | 0.161 |
| 2011 | 0.566 | 0.720 | 0.156 |
| 2012 | 0.567 | 0.723 | 0.151 |
| 2013 | 0.566 | 0.745 | 0.136 |
| 2014 | 0.555 | 0.775 | 0.134 |
| 2015 | 0.485 | 0.695 | 0.137 |
This data on its own does not provide much information (aside from the fact that data on undernourishment is not available before 1991) – the global levels of democracy and women’s empowerment have generally increased at similar rates over time while overall levels of undernourishment have decreased at a slower rate. Correlations between these variables are unclear.
Examining the data by country, we use the same commands to group by country then generate summary statistics.
We can do the same thing to look at cross-state comparisons:
vdem_un_country <- group_by(vdem_un, country_name)
vdem_un_country <- summarize(vdem_un_country, demmean=mean(v2x_polyarchy, na.rm=TRUE),
gendermean=mean(v2x_gender, na.rm=TRUE),
nourmean=mean(undernour, na.rm=TRUE))
kable(vdem_un_country, digits=3)| country_name | demmean | gendermean | nourmean |
|---|---|---|---|
| Afghanistan | 0.181 | 0.269 | 0.358 |
| Albania | 0.320 | 0.529 | NaN |
| Algeria | 0.243 | 0.394 | 0.070 |
| Angola | 0.102 | 0.472 | 0.440 |
| Argentina | 0.612 | 0.739 | 0.050 |
| Armenia | 0.453 | 0.642 | 0.153 |
| Australia | 0.898 | 0.791 | NaN |
| Austria | 0.881 | 0.830 | NaN |
| Azerbaijan | 0.252 | 0.477 | 0.141 |
| Bangladesh | 0.437 | 0.580 | 0.238 |
| Barbados | 0.584 | 0.823 | 0.053 |
| Belarus | 0.346 | 0.810 | NaN |
| Belgium | 0.815 | 0.852 | NaN |
| Benin | 0.403 | 0.697 | 0.190 |
| Bhutan | 0.127 | 0.516 | NaN |
| Bolivia | 0.479 | 0.519 | 0.300 |
| Bosnia and Herzegovina | 0.340 | 0.765 | NaN |
| Botswana | 0.602 | 0.544 | 0.309 |
| Brazil | 0.596 | 0.597 | 0.090 |
| Bulgaria | 0.421 | 0.771 | NaN |
| Burkina Faso | 0.405 | 0.705 | 0.238 |
| Burma_Myanmar | 0.150 | 0.213 | 0.406 |
| Burundi | 0.205 | 0.442 | NaN |
| Cambodia | 0.251 | 0.455 | 0.243 |
| Cameroon | 0.256 | 0.480 | 0.255 |
| Canada | 0.874 | 0.860 | NaN |
| Cape Verde | 0.477 | 0.683 | 0.153 |
| Central African Republic | 0.228 | 0.412 | 0.428 |
| Chad | 0.200 | 0.359 | 0.448 |
| Chile | 0.585 | 0.662 | 0.055 |
| China | 0.097 | 0.571 | 0.163 |
| Colombia | 0.467 | 0.538 | 0.104 |
| Comoros | 0.331 | 0.484 | NaN |
| Congo_Republic of the | 0.223 | 0.343 | 0.363 |
| Costa Rica | 0.867 | 0.782 | 0.055 |
| Croatia | 0.668 | 0.838 | NaN |
| Cuba | 0.108 | 0.617 | 0.082 |
| Cyprus | 0.661 | 0.636 | NaN |
| Czech Republic | 0.500 | 0.750 | NaN |
| Denmark | 0.920 | 0.926 | NaN |
| Djibouti | 0.201 | 0.250 | 0.487 |
| Dominican Republic | 0.506 | 0.692 | 0.250 |
| Ecuador | 0.548 | 0.527 | 0.170 |
| Egypt | 0.224 | 0.420 | 0.050 |
| El Salvador | 0.358 | 0.398 | 0.127 |
| Eritrea | 0.084 | 0.335 | NaN |
| Estonia | 0.911 | 0.904 | NaN |
| Ethiopia | 0.171 | 0.288 | 0.523 |
| Fiji | 0.434 | 0.571 | 0.052 |
| Finland | 0.887 | 0.911 | NaN |
| France | 0.893 | 0.823 | NaN |
| Gabon | 0.286 | 0.478 | 0.065 |
| Gambia | 0.430 | 0.487 | 0.137 |
| Georgia | 0.498 | 0.748 | 0.214 |
| Germany | 0.771 | 0.861 | NaN |
| Ghana | 0.438 | 0.661 | 0.163 |
| Greece | 0.693 | 0.773 | NaN |
| Guatemala | 0.342 | 0.288 | 0.170 |
| Guinea | 0.209 | 0.387 | 0.232 |
| Guinea-Bissau | 0.208 | 0.444 | 0.244 |
| Guyana | 0.442 | 0.764 | 0.129 |
| Haiti | 0.265 | 0.462 | 0.570 |
| Honduras | 0.395 | 0.492 | 0.181 |
| Hungary | 0.434 | 0.676 | NaN |
| Iceland | 0.878 | 0.802 | NaN |
| India | 0.707 | 0.574 | 0.188 |
| Indonesia | 0.348 | 0.443 | 0.158 |
| Iran | 0.192 | 0.296 | 0.057 |
| Iraq | 0.174 | 0.401 | 0.228 |
| Ireland | 0.866 | 0.755 | NaN |
| Israel | 0.733 | 0.739 | NaN |
| Italy | 0.815 | 0.753 | NaN |
| Ivory Coast | 0.304 | 0.483 | 0.136 |
| Jamaica | 0.554 | 0.819 | 0.081 |
| Japan | 0.874 | 0.676 | NaN |
| Jordan | 0.189 | 0.397 | 0.059 |
| Kazakhstan | 0.295 | 0.674 | 0.050 |
| Kenya | 0.315 | 0.373 | 0.301 |
| Korea_North | 0.092 | 0.278 | 0.354 |
| Korea_South | 0.534 | 0.592 | 0.050 |
| Kyrgyzstan | 0.352 | 0.643 | 0.120 |
| Laos | 0.123 | 0.317 | 0.346 |
| Latvia | 0.866 | 0.934 | NaN |
| Lebanon | 0.437 | 0.477 | 0.050 |
| Lesotho | 0.286 | 0.521 | 0.131 |
| Liberia | 0.306 | 0.456 | 0.363 |
| Libya | 0.117 | 0.239 | NaN |
| Lithuania | 0.868 | 0.917 | NaN |
| Macedonia | 0.529 | 0.719 | NaN |
| Madagascar | 0.331 | 0.514 | 0.335 |
| Malawi | 0.292 | 0.353 | 0.302 |
| Malaysia | 0.311 | 0.565 | 0.050 |
| Maldives | 0.223 | 0.527 | 0.111 |
| Mali | 0.376 | 0.581 | 0.123 |
| Mauritania | 0.271 | 0.413 | 0.115 |
| Mauritius | 0.730 | 0.590 | 0.063 |
| Mexico | 0.449 | 0.530 | 0.055 |
| Moldova | 0.603 | 0.781 | NaN |
| Mongolia | 0.440 | 0.636 | 0.347 |
| Montenegro | 0.516 | 0.797 | NaN |
| Morocco | 0.210 | 0.449 | 0.060 |
| Mozambique | 0.241 | 0.538 | 0.409 |
| Namibia | 0.346 | 0.409 | 0.340 |
| Nepal | 0.236 | 0.329 | 0.177 |
| Netherlands | 0.894 | 0.870 | NaN |
| New Zealand | 0.880 | 0.841 | NaN |
| Nicaragua | 0.411 | 0.534 | 0.332 |
| Niger | 0.365 | 0.549 | 0.216 |
| Nigeria | 0.312 | 0.588 | 0.096 |
| Norway | 0.905 | 0.922 | NaN |
| Pakistan | 0.293 | 0.469 | 0.231 |
| Panama | 0.426 | 0.577 | 0.227 |
| Papua New Guinea | 0.410 | 0.407 | NaN |
| Paraguay | 0.357 | 0.382 | 0.136 |
| Peru | 0.506 | 0.623 | 0.198 |
| Philippines | 0.448 | 0.714 | 0.191 |
| Poland | 0.519 | 0.772 | NaN |
| Portugal | 0.684 | 0.692 | NaN |
| Qatar | 0.041 | 0.297 | NaN |
| Russia | 0.259 | 0.602 | NaN |
| Rwanda | 0.217 | 0.518 | 0.491 |
| Sao Tome and Principe | 0.331 | 0.677 | 0.166 |
| Saudi Arabia | 0.020 | 0.145 | 0.050 |
| Senegal | 0.562 | 0.653 | 0.239 |
| Serbia | 0.297 | 0.705 | NaN |
| Seychelles | 0.337 | 0.654 | NaN |
| Sierra Leone | 0.325 | 0.367 | 0.361 |
| Slovakia | 0.734 | 0.873 | NaN |
| Slovenia | 0.810 | 0.882 | NaN |
| Solomon Islands | 0.490 | 0.352 | 0.147 |
| Somalia | 0.191 | 0.240 | NaN |
| South Africa | 0.389 | 0.448 | 0.051 |
| South Sudan | 0.219 | 0.496 | NaN |
| Spain | 0.651 | 0.703 | NaN |
| Sri Lanka | 0.578 | 0.616 | 0.289 |
| Sudan | 0.203 | 0.284 | NaN |
| Suriname | 0.658 | 0.694 | 0.120 |
| Swaziland | 0.118 | 0.382 | 0.201 |
| Sweden | 0.893 | 0.919 | NaN |
| Switzerland | 0.867 | 0.839 | NaN |
| Syria | 0.155 | 0.439 | NaN |
| Tajikistan | 0.248 | 0.523 | 0.362 |
| Tanzania | 0.356 | 0.659 | 0.337 |
| Thailand | 0.327 | 0.628 | 0.174 |
| Togo | 0.251 | 0.574 | 0.289 |
| Trinidad and Tobago | 0.655 | 0.760 | 0.125 |
| Tunisia | 0.222 | 0.593 | 0.050 |
| Turkey | 0.529 | 0.466 | 0.050 |
| Turkmenistan | 0.153 | 0.385 | 0.074 |
| Uganda | 0.264 | 0.586 | 0.257 |
| Ukraine | 0.508 | 0.788 | NaN |
| United Kingdom | 0.873 | 0.769 | NaN |
| United States | 0.838 | 0.793 | NaN |
| Uruguay | 0.699 | 0.751 | 0.053 |
| Uzbekistan | 0.177 | 0.420 | 0.089 |
| Vanuatu | 0.480 | 0.756 | 0.084 |
| Venezuela | 0.716 | 0.713 | 0.122 |
| Vietnam_Democratic Republic of | 0.194 | 0.599 | 0.267 |
| Yemen | 0.194 | 0.192 | 0.286 |
| Zambia | 0.377 | 0.534 | 0.444 |
| Zimbabwe | 0.258 | 0.363 | 0.406 |
Breaking the data down by country shows us that while the average levels of democracy and gender empowerment in a country are often equivalent, in some countries (particularly current or former socialist countries) they diverge (general political rights may be limited, but women have higher degrees of equality within them). The correlation with undernourishment is unclear, except to note that this data is missing for many countries, notably high income countries. As with the annual data, levels of women’s empowerment are generally higher than levels of democracy.
Part 3: Graphics
1. Scatterplot by year
To better understand the correlation between electoral democracy (v2x_polyarchy) and undernourishment (undernour), I will graph the data. First I will create a smaller version of the dataset covering only the relevant years (1991-2015) then I will create a grid of scatterplots by year, with best fit lines and 95% confidence intervals.
dem_nour <- filter(vdem_un, year>1990)
g1 <- ggplot(dem_nour, aes(x=v2x_polyarchy, y=undernour)) +
geom_point() +
geom_smooth(method="lm") +
facet_wrap( ~ year)
g1
These graphs show us that for the limited number of countries for which data is available, the weak correlation between regime type and undernourishment has decreased over time. While in the early 1990s there was a weak but clear correlation between democracy score and percent undernourished, by 2015 this relationship disappeared. This correlation might appear stronger, however, if high income countries were included in the available data. In addition, it appears that the overall number of countries for which data is available decreases over time as well.
2. Scatterplot by country
I will now graph the same data by country instead of by year.
g2 <- ggplot(dem_nour, aes(x=v2x_polyarchy, y=undernour)) +
geom_point() +
geom_smooth(method="lm") +
facet_wrap( ~ country_name, ncol=5)
g2
These graphs show us that for most countries, the level of undernourishment does not change with the level of democracy – the level of undernourishment remains fairly constant even when there is a change in the level of democracy. There are a few countries that are exceptions to this trend – countries such as Angola and Georgia experienced lower levels of malnourishment when they had higher levels of electoral democracy, while a handful of countries (Mongolia, Zambia) experienced slighly higher levels of malnourishment with higher levels of electoral democracy.
3. Time series line plot
To show change over time for these a variables within an individual countries, I create time series graphs. I have chosen Angola, Mongolia, and Thailand as sample countries because these three countries demonstrate three different trends in the cross-national data for democracy and undernourishment. These graphs will should change in electoral democracy score.
graphdata <- filter(vdem_un, country_name=="Angola" | country_name=="Mongolia" | country_name=="Thailand")
g3 <- ggplot(graphdata, aes(x=year, y=v2x_polyarchy, group=country_name, color=country_name)) +
geom_line()
g3
This graph shows the variations in types of political transitions around the world. Some countries (such as Mongolia) make a stable transition from an authoritarian system to a democratic system. Others, such as Thailand, have experienced a series of coups that have moved the country repeatedly between democracy and non-democracy. Other countries like Angola may experience transitions (such as the 1991 accords that were designed to create free and fairr elections) but remain autocracies (in this case, due to civil war).
4. Time series graph with plotly
I recreate the above graph using the “plotly”" package.
plot_ly(graphdata, x = ~year, y = ~v2x_polyarchy, color = ~country_name, type = "scatter", mode = "lines")Plotly produces the same graph, but allows you to view the exact data along the line with an interactive feature.
5. Motion graph
With GoogleVis, I can graph the change in malnourishment and democracy over time in my three sample countries. This graph takes a long time to load, but it will appear in a separate website when you run it.
M1 <- gvisMotionChart(graphdata, idvar = "country_name", timevar = "year", xvar = "v2x_polyarchy", yvar = "undernour", colorvar = "country_name", options=list(width=600, height=400))
plot(M1)## starting httpd help server ...## doneprint(M1, "chart")6. Interactive map
GoogleVis can also plot a world map with color variations based on the level of democracy in a country. To develop this graph, I first select one year for analysis (in this case, 2010). On this graph, darker shades of green correspond to higher levels of democracy.
data2010 <- filter(vdem_un, year==2010)
M2 <- gvisGeoChart(data2010, locationvar = "country_name", colorvar = "v2x_polyarchy", options=list(width=600, height=400))
plot(M2)
print(M2, "chart")7. 3D scatterplot
Finally, it is possible to graph both the democracy score and level of malnourishment for my three sample countries on the same 3D graph.
plot_ly(graphdata, x = ~year, y = ~v2x_polyarchy, z = ~undernour,
type = "scatter3d", color = ~country_name)## No scatter3d mode specifed:
## Setting the mode to markers
## Read more about this attribute -> https://plot.ly/r/reference/#scatter-modeWhile the relationship for Thailand is unclear, it is possible to that both Angola and Thailand’s levels of undernourishment have decreased over time, despite a lack of significant change in their democracy score during the time period under consideration.