The Correlation of the Number of Refugees to the Occurrence of Conflicts and Civil Wars - Assignment Example

Short assignment IV Select two appropriate variables from any data set you prefer to work with and calculate Spearman’s rho. Briefly discuss your results. (Max. 250 words) Spearmans Rank Correlation is a technique used to test the direction and strength of the relationship between two variables. In other words, it is a device to show whether any one set of numbers has an effect on another set of numbers. The statistic falls between -1 and +1. To Spearman’s rho, let us consider the following two variables: 1. Hostility Level (D8) - 1 no militarized threat 2threat to use force 3display of force 4use of force 5war/ 2. Geographic Region (D12) - 1 Nth America 2 Central & S Amer 3 Africa 4 SW Asia 5 E Asia & Pacific 6 Middle East 7 Europe/ Using SPSS, the following correlation matrix is obtained. Correlations Hostility Level Geographic Region Spearmans rho Hostility Level Correlation Coefficient 1.000 -.047(**) Sig. (2-tailed) . .008 N 3207 3207 Geographic Region Correlation Coefficient -.047(**) 1.000 Sig. (2-tailed) .008 . N 3207 3207 ** Correlation is significant at the 0.01 level (2-tailed). We find that a very low negative correlation coefficient is obtained. That is there is a weak negative correlation between the two variables. It is found to be significant at 99% level of significance. This means that when the level of one variable increases, the level of the other variable tends to decrease. This can be further described as: The hostility level may be higher or highest in North America leading to war, while the hostility level may be the least in Europe having no militarized threat. 2)      Briefly comment upon options to compare means when sample sizes are small and assumptions of normality or equal variances are not met. (Max. 250 words) The two sample t-test is used to compare the sample means when the sample sizes are small and the assumptions of normality or equal variances are met. However, in the other case when the assumptions are not the following non-parametric tests are used: Mann-Whitney test and Kolmogorov-Smirnov test (KS-test). These tests are called the non-parametric tests. As both these test do not rest on any assumption concerning the underlying distributions, they are more widely applicable than the t-test. So whenever there exist some doubts about the validity of these assumptions, the Mann-Whitney test is an excellent alternative. There is a clear similarity between the t test on independent samples and the non-parametric tests as both are testing the identity of two independent populations.     * The t test on independent samples follows the assumptions that the two populations are normal with identical variances is essential (so, in fact, the t test is testing the identity of two populations).     * The Mann-Whitney test, on the other hand, is non parametric : it does not rest on any assumption concerning the underlying distributions. It is therefore more widely applicable than the t-test. Even when the assumptions of the t test are justified, it can be shown that the Mann-Whitney is hardly less powerful that the t test for large samples. * The Kolmogorov-Smirnov test (KS-test) tries to determine if two datasets differ significantly. The KS-test has the advantage of making no assumption about the distribution of data.   2) Choose appropriate data and conduct a binary logistic regression with a quantitative independent variable. Briefly discuss your results. (Max. 300 words). A binary logistic regression is used to perform logistic regression on a binary response variable. A binary variable only has two possible values, such as presence or absence of a particular disease. Let us consider the following variables: 1. Reciprocity (D9) - 0no 1yes/ ---- this is the response variable with two levels 2. Fatalities (raw) (D5a) --- This is the quantitative variable 3. Geographic Region (D12) - 1 Nth America 2 Central & S Amer 3 Africa 4 SW Asia 5 E Asia & Pacific 6 Middle East 7 Europe/--- This is another categorical variable. Because the responseReciprocity is categorized into yes and no, a binary logistic regression analysis is appropriate to investigate the effects of fatalities and geographic region on reciprocity. Binary Logistic Regression: Reciprocity versus Fatalities, Geographic region Link Function: Logit Response Information Variable Value Count Reciprocity yes 3134 (Event) no 73 Total 3207 Factor Information Factor Levels Values Geographic region 6 Africa, Central & S Amer, E Asia & Pacific, Europe, Middle East, SW Asia Logistic Regression Table Odds 95% CI Predictor Coef SE Coef Z P Ratio Lower Upper Constant 2.81655 0.368149 7.65 0.000 Fatalities 0.0012216 0.0002956 4.13 0.000 1.00 1.00 1.00 Geographic region Central & S Amer -0.117999 0.554822 -0.21 0.832 0.89 0.30 2.64 E Asia & Pacific -1.20536 0.451890 -2.67 0.008 0.30 0.12 0.73 Europe 1.22611 1.06992 1.15 0.252 3.41 0.42 27.75 Middle East -1.25829 0.400459 -3.14 0.002 0.28 0.13 0.62 SW Asia 0.267319 0.807148 0.33 0.741 1.31 0.27 6.36 Log-Likelihood = -233.566 Test that all slopes are zero: G = 229.458, DF = 6, P-Value = 0.000 Goodness-of-Fit Tests Method Chi-Square DF P Hosmer-Lemeshow 2.545 8 0.960 Table of Observed and Expected Frequencies: (See Hosmer-Lemeshow Test for the Pearson Chi-Square Statistic) Group Value 1 2 3 4 5 6 7 8 9 10 yes Obs 282 308 421 340 330 372 352 328 325 76 Exp 282.2 304.8 424.0 340.0 330.0 372.0 352.0 328.0 325.0 76.0 no Obs 51 14 8 0 0 0 0 0 0 0 Exp 50.8 17.2 5.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Total 333 322 429 340 330 372 352 328 325 76 Value Total yes Obs 3134 Exp no Obs 73 Exp Total 3207 From the logistic regression table, we find that the odds ratio of fatalities is equal to 1. This means that Reciprocity is unaffected by one unit increase in fatalities. Further the G value (229.458) with p-value < 0.000, shows that one of the coefficients of the variables under study is not equal to 0. The Hosmer-Lemeshow Goodness of fit test provides a highly insignificant p-value (0.960). This means that there is insufficient evidence to claim that the model does not fit the data. Table of Observed and Expected Frequencies - allows us to see how well the model fits the data by comparing the observed and expected frequencies. There is insufficient evidence that the model does not fit the data well, as the observed and expected frequencies are similar. This supports the conclusions made by the Goodness of Fit Test. 4)      Briefly discuss the article by Salehyan, Idean and Kristian Skrede Gleditsch (2006): ‘Refugees and the Spread of Civil War’, International Organization, 60, 335-366, notably as regards statistical techniques used and results obtained. (Max. 300 words). The increase in political conflicts and violence leading to wars has forced us to analyse and examine every possible aspect disturbing the peace process of one country and its neighbouring countries. The one factor that has been paid least attention is the refugee flow. Though these people do not contribute to violence in vivid manner, they cannot be dispelled from the purpose even. The flow of refugees from insecure camps also leads to spread of epidemic diseases. The security of the host country is also adversely affected. A statistical analysis of the link between refugees and civil conflict since mid-twentieth century has shown that the presence of refugees from neighbouring countries is a crucial reason for conflict diffusion. The data under study is from the Uppsala/PRIO Conflict Data Set. The main dependent variable was conflict onset, coded 1 for the first year of conflict and 0 for no conflict. A chi-square/cross-tabulation between conflict onset and refugee flow from neighbouring states shows that the relationship between conflict and refugees in not a deterministic one. This may be due to the security measures taken by the host countries through various organisations. Further there are several other factors that also contribute to the risk of civil war. Hence these factors (conflicts in neighbouring countries, transborder ethnic groups, effect of income measured by GDP, impact of democracy based on Polity 4, ethnic relations and total population) are considered as control measures and suitable models are developed. Similarly models are developed by adding significant variables and the corresponding p-values are analyzed. From all the models, we find that there is sufficient evidence to support the null hypothesis that refugees significantly contribute to the occurrence of conflicts and civil wars. Though it is also shown that refugees do not involve in violence, the occurrence can be prevented by political intervention of the individual government and the United Nations. Sources: 1. http://www.revision-notes.co.uk/revision/181.html 2. http://www.physics.csbsju.edu/stats/KS-test.html 3. http://faculty.chass.ncsu.edu/garson/PA765/logistic.htm Read More

Spearman's Rank Correlation is a technique used to test the direction and strength of the relationship between two variables. In other words, it is a device to show whether any one set of numbers has an effect on another set of numbers. The statistic falls between -1 and +1.

To Spearman’s rho, let us consider the following two variables:

1. Hostility Level (D8) - 1 'no militarized threat' 2'threat to use force' 3'display of force' 4'use of force' 5'war'/
2. Geographic Region (D12) - 1 'Nth America' 2 'Central & S Amer' 3 'Africa' 4 'SW Asia' 5 'E Asia & Pacific' 6 'Middle East' 7 'Europe'/

Using SPSS, the following correlation matrix is obtained.

                                                             Correlations

 

 

 

Hostility Level

Geographic Region

Spearman's rho

Hostility Level

Correlation Coefficient

1.000

-.047(**)

 

 

Sig. (2-tailed)

.

.008

 

 

N

3207

3207

 

Geographic Region

Correlation Coefficient

-.047(**)

1.000

 

 

Sig. (2-tailed)

.008

.

 

 

N

3207

3207

**  Correlation is significant at the 0.01 level (2-tailed).

We find that a very low negative correlation coefficient is obtained. That is there is a weak negative correlation between the two variables. It is found to be significant at a 99% level of significance.

This means that when the level of one variable increases, the level of the other variable tends to decrease. This can be further described as the hostility level may be higher or highest in North America leading to war, while the hostility level may be the least in Europe having no militarized threat.

2)      Briefly comment upon options to compare means when sample sizes are small and assumptions of normality or equal variances are not met. (Max. 250 words)

 The two-sample t-test is used to compare the sample means when the sample sizes are small and the assumptions of normality or equal variances are met. However, in the other case when the assumptions are not the following non-parametric tests are used: Mann-Whitney test and Kolmogorov-Smirnov test (KS-test). These tests are called non-parametric tests.

As both these tests do not rest on any assumption concerning the underlying distributions, they are more widely applicable than the t-test. So whenever there exist some doubts about the validity of these assumptions, the Mann-Whitney test is an excellent alternative.

There is a clear similarity between the t-test on independent samples and the non-parametric tests as both are testing the identity of two independent populations.

    * The t-test on independent samples follows the assumptions that the two populations are normal with identical variances is essential (so, in fact, the t-test is testing the identity of two populations).

    * The Mann-Whitney test, on the other hand, is non-parametric: it does not rest on any assumption concerning the underlying distributions. It is therefore more widely applicable than the t-test. Even when the assumptions of the t-test are justified, it can be shown that the Mann-Whitney is hardly less powerful than the t-test for large samples.

    * The Kolmogorov-Smirnov test (KS-test) tries to determine if two datasets differ significantly. The KS-test has the advantage of making no assumption about the distribution of data.  

• Choose appropriate data and conduct a binary logistic regression with a quantitative independent variable. Briefly discuss your results. (Max. 300 words).

A binary logistic regression is used to perform logistic regression on a binary response variable. A binary variable only has two possible values, such as the presence or absence of a particular disease.

Let us consider the following variables:

1. Reciprocity (D9) - 0'no' 1'yes'/ ---- this is the response variable with two levels
2. Fatalities (raw) (D5a) --- This is the quantitative variable                                               

3. Geographic Region (D12) - 1 'Nth America' 2 'Central & S Amer' 3 'Africa' 4 'SW Asia' 5 'E Asia & Pacific' 6 'Middle East' 7 'Europe'/--- This is another categorical variable.

Because the response- Reciprocity is categorized into yes and no, a binary logistic regression analysis is appropriate to investigate the effects of fatalities and geographic region on reciprocity.

 

Read More