Method:
A linear regression model was used to find how well suicide rates can be computed with factors such as
GDP Growth Rate, GDP Per Capita, Inflation Rate and Unemployment Rate. Before proceeding with
running the checks and creating the model, the data was filtered to remove the years for which suicide
mortality rate couldn’t be found. Additionally, the data was normalized so that all the values for the
different independent and dependent variables fall in the same range.
For reference, check the excel sheet below to get an overview of original, filtered and normalized data.
Model Analysis
The model was run such that Suicide Rate was used as the dependent variable and all the independent
variables- GDP Growth Rate, GDP Per Capita, Inflation Rate and Unemployment Rate were used.
Checks for Validity of the Modal:
Checking for Linearity :For checking the linearity, the residuals were plotted against the fitted values.
As the plot shows the data doesn’t seem to be linear but given that the data points are very few, it could
be contributor of this lack of linearity.
Checking for Homoscedasticity: This allows to check if the variance is of errors is constant across all the
independent variables.
The plot shows that the square root of standardized residuals is uniformly distributed across the fitted
values. However, given that the number of points is few, we also perform a non-constant variance test.
The non constant variance test gives a p value of 0.62 which is quite high. Hence, we fail to
reject the null hypothesis indicating that variance is constant and hence there is
homoscedasticity.
Checking for Normality: In order for the model to be valid, it should follow a normal
distribution. The Q-Q plot and Shapiro test are used to check for these.
The Q-Q plot gives a straight line which indicates normality of data.
The High p- value of the Shapiro test shows that we fail to reject null hypothesis and the data is
normal .
Auto Correlation Matrices:
These matrices show how the different variables are correlated with each other. From this matrix
it can be seen that the value of correlation for independent variables are not strongly correlated
with each. The values are below or close to 0.5 which shows weak correlation.
This matrix also shows that the value of suicide rate is strongly correlated with GDP Per Capita
which give a preemptive idea that the model which includes this value should perform well.
To test Autocorrelation as Durbin Watson Test is also performed.
The Autocorrelation value from the test is lower than 0.5 which shows that there isn’t ,the presence of
autocorrelation in the data.
The Model:
A linear regression model is created which includes all the independent variables and the suicide rate as
the dependent variable. The summary is given below:
Model 2:
Form the correlation matrix, there seems to be somewhat of a correlation between GDP Growth Rate
and Unemployment Rate. A model without the unemployment rate variable is created and test. The
summary is given below.
In order to test if unemployment rate is really necessary, we run an ANOVA test comparing the model
without unemployment and with unemployment rate.
The p value for the ANOVA is high indicating that we fail to reject the null hypothesis.
Thus, The ANOVA shows that the model with the unemployment rate is not significantly better than
without the unemployment rate.
Finally, to test, if the model with unemployment rate is indeed better or worse than without
unemployment rate, an AIC test is performed. The lower the value of the AIC test, the better the model.
A lower value of AIC test is obtained in the model without unemployment rate indicating that this is
better