#passion driven statistics

[statistiken]

Final comments

and it's done.

Thank you Lisa Dierker and the team of Wesleyan University. I did enjoy this course, my first from coursera.

I was able to at least experience (somewhat program in) SAS! 

And of course to very helpful tumblrs (tumblr users?) and other students.  n

[statistiken]

PASSION DRIVEN STATISTICS: FINAL PROJECT

Oil reserves of the world becomes more and more scare. The hungriest countries that need it most will have to make a move to protect and strengthen their hold of the world's remaining oil reserves. One way to do this is to become a consistent world power, esp. in military.  Furthermore, the least democratic (autocratic in that sense) countries' leaders may have the greatest hold and control of its spending without having to face the people's usual non-conformity with military.

The study looks at the countries of different amounts of oil consumption and if that may relate to perceive amounts of possible military action to protect that level of consumption. It also take into consideration different democracy level and how that could affect the association.

Sample

Total number of matching records (n=56) of the 3 variables used: amrmedforcesrate, oilperperson, polityscore were from the Gapminder dataset.

Gapminder [http://www.gapminder.org] is a non-profit organization promoting sustainable global development and achievement of the United Nations Millennium Development Goals.

Measures

polityscore: 2009 Democracy score (polity); A measure calculated by subtracting an autocracy score from a democracy score. The summary measure of a country's democratic and free nature. -10 is the lowest value, 10 is the highest sourced by Polity IV Project.

armedforcesrate: Armed forces personnel (% of total labor force) sourced by Work Development Indicators

oilperperson: 2010 oil Consumption per capita in tonnes per year and person sourced by BP.

Univariate

The average armed forces personnel is only 1.264% of the total labor force.

The global oil consumption per capita is 1.7 tonnes/year.

An average of 3.716 for the polity score and a standard deviation of 6.304 shows a negative skewness of -0.733 for this variable. 

Figure 0: Armed forces rate grouped in Quartiles in comparison with the oil consumption variable.

Bivariate

A Pearson correlation of coefficient analysis showed that the oil consumption per capita (quantitative explanatory variable) is positively and significantly associated with the armed forces personnel's percent in the total labor force (quantitative response variable); r = 0.52156, p< 0.0001. That is, the more a country consumes oil,  the more likely the country is to increase its armed forces personnel. 

The polity score of a country did moderate the correlation of coefficient's significance. It shows that that the countries with very low polity score (Figure 1) maintains the significant correlation while the countries with a low polity score (Figure 2) and the countries with a high polity score (Figure 3) shows no statistically significant correlation between the armed forces rate and oil consumption per capita variables.

Figure 1: A scatter plot of the association of oil consumption per capita to the armed forces personnel percentage, grouped by poligroup1 [countries with polity score between -10 to -1] that has Pearson's coefficient correlation(r) of 0.6612 with p>.0101.

 Figure 2: A scatter plot of the association of oil consumption per capita to the armed forces personnel percentage, grouped by poligroup2 [countries with polity score between 1 to 5] that has Pearson's coefficient correlation(r) of .34072 with a non-significant p value.

Figure 3: A scatter plot of the association of oil consumption per capita to the armed forces personnel percentage, grouped by poligroup3 [countries with polity score between 6 to 10] that has Pearson's coefficient correlation(r) of -.10423 with a non-significant p value.

[workingwomenstats]

Final: What Drives Women in the Workforce?

The association between female employment rate and the urban rate, democratization

Emily McCormick, Nashville, Tennessee, U.S.A.

INTRODUCTION

The rights of women, including strides made in the workforce, have rapidly changed within the past century, yet the rights of women to work differ around the world. In what types of countries does female employment thrive?

Does a country with a more urban population also have more women in the workforce?

Does a country with a higher polity score, that is, a higher level of democratization, also have more women in the workforce?

METHODOLOGY

Data grouped as follows:

Female employment rate was grouped as FemGroup1 (less than 30%), FemGroup2 (30%-45%), FemGroup3 (45%-60%) and FemGroup4 (greater than 60%). There were 111 null values, representing 38 percent of the data set.

Urban rate was grouped as Urban1 (less than 30%), Urban2 (30%-45%), Urban3 (45%-60%), Urban 4 (60%-75%) and Urban5 (greater than 75%). There were 81 null values, representing 28 percent of the data set.

As the polity score is a score that could be considered a categorical variable, this data was not grouped in any way.

The univariate procedure was run for female employment rate, urban rate and polity score:

The mean for the female employment rate is 47.55%, with a standard deviation of 14.63.

The mean for the urban rate is 57.14%, with a standard deviation of 24.02.

The mean for the polity score is 3.72, with a standard deviation of 6.3.

RESULTS

First Hypothesis:

H0: The female employment rate has no association with the urban rate.

H1: There is an association between the female employment rate and the urban rate.

As this hypothesis requires the analysis of two quantitative variables, the data was examined using the Pearson Correlation Co-Efficient. Is there be a strong correlation or a weak correlation between the two variables, female employment and the urban rate?

This was the Pearson Correlation procedure for my two data-managed variables, Urban and FemGroup:

The correlation co-efficient is just over 0.41, with a p-value of less than .0001, making the relationship statistically significant. With a value of 0.41, which is closer to 0 than it is to 1, this does not indicate a strong linear relationship.

To ensure that I did not overly-manage my data, I ran the Pearson Correlation for the original, ungrouped data: the urban rate and female employment rate.

Again, the p-value is .0001, and the correlation co-efficient is even closer to zero, at -0.30298; again, this does not indicate a strong linear relationship.

The scatter plot supports the conclusion that the female employment rate is not associated with the urban rate:

Second Hypothesis:

H0: The female employment rate has no association with the polity score.

H1: There is an association between the female employment rate and the polity score.

As this hypothesis requires the analysis of a quantitative and categorical variable, the data was examined using the Analysis of Variance (ANOVA). Will there be a strong correlation or a weak correlation between the two variables, female employment and the polity score?

With a p-value of 0.068, along with an examination of the means of female employment at each level of polity score, there does not seem to be a relationship - negative or positive - between female employment rate and the polity score.

A look at a box plot graph confirms that there is no association between the female employment rate and the polity score:

The ANOVA procedure requires a post-hoc test. The DUNCAN post-hoc test further confirms the lack of an association between the two variables, female employment rate and polity score.

DISCUSSION

Despite my expectations, there does not seem to be an association between urban rate and female employment nor between the polity score and female employment.

The Gapminder set suffers from many gaps in the data, making a good analysis of these factors difficult.

Female employment is a complex puzzle. Many women make the choice to stay at home, which could have little to do with where they live or how democratic the country may be. Perhaps a better indicator would be income per gender, to measure how much women are earning, or perhaps determining a level of employment, measuring blue collar versus white collar among men and women. Education levels could also contribute to a study of the progress women make in the workforce.

SAS Code:

libname mydata "/courses/u_coursera.org1/i_1006328/c_5333" access=readonly; DATA new; set mydata.gapminder_pds; IF FEMALEEMPLOYRATE= . THEN FemGroup=0; /*Category for missing data*/ ELSE IF FEMALEEMPLOYRATE LE 30.00 THEN FemGroup=1;/*LowFemEmployment*/ ELSE IF FEMALEEMPLOYRATE LE 45.00 THEN FemGroup=2;/*MidFemEmployment*/ ELSE IF FEMALEEMPLOYRATE LE 60.00 THEN FemGroup=3; /*MidHighFemEmployment*/ ELSE IF FEMALEEMPLOYRATE GT 60.00 THEN FemGroup=4; /*HighFemEmployment*/ IF urbanrate= . THEN Urban=0; /*Category for missing data*/ ELSE IF urbanrate LE 30.00 THEN Urban=1;/*LowUrbanEmployment*/ ELSE IF urbanrate LE 45.00 THEN Urban=2;/*MidLowUrbanEmployment*/ ELSE IF urbanrate LE 60.00 THEN Urban=3; /*MidUrbanEmployment*/ ELSE IF urbanrate LE 75.00 THEN Urban=4; /*MidHighUrbanEmployment*/ ELSE IF urbanrate GT 75.00 THEN Urban=5; /*HighUrbanEmployment*/ PROC SORT; by Country; /*PROC CORR; var urbanrate femaleemployrate;*/ PROC ANOVA; CLASS polityscore; MODEL FEMALEEMPLOYRATE=polityscore; MEANS polityscore/DUNCAN; PROC FREQ; TABLES FemGroup Urban polityscore; PROC Univariate; var FEMALEEMPLOYRATE; PROC Univariate; var urbanrate; Proc MEANS; var FemGroup urban; run;

[jpassionstats]

Assignment 5: Association between Variables

INSTRUCTIONS:   Use the SAS graphing wizard to visualize an association between two variables. Make sure to follow the Graphing Decisions Flowchart to help you decide what type of graph to use and if any variables need to be managed first. Submit the Log that is generated automatically. Also note that the Univariate Procedure may help to describe any quantitative variables you choose.

To explore the association between alcohol consumption and suicide rate, I have created a bar chart where I look at the level of suicide rate for each alcohol consumption group (using the categorical variable previously created). Below is the bar chart:

We can see how the aveerage value of suicide rate increases with alcohol consumption group, with a peak for group 6 (the highest level of alcohol consumption). There seems to be a positve relationship between the two variables. Alcohol group six contains the 6.59% of observation (11 observations).

Below is the scatterplot of the two variables:

The linear regression confirm the postive relationship: as the alcohol consumption grows, the suicide rate per 100,000 grows as well. The two lines above and below the linear regression represent the 95% confidence interval: we can see that almost all the observations fall within the confidence interval.

We'll see in the next post whether the difference in suicide rate between each alcohol group is significant.

[jpassionstats]

Assignment 2 & 3: Frequency Tables

Assignment: Write a SAS program that calls in your chosen data set (AddHealth, NESARC, Gapminder or Mars Study) and outputs frequency tables. If you have decided you are only interested in a subset of the data, please include the appropriate commands. Also make sure to provide comments that explain what each part of the program does.

I begin the exploration of the variables through my first SAS program. I've got a bit of experience with R, and at first sight SAS syntax seems much easier (at least to peform basic tasks).

Basically all the variables in the Gapminder data set are quantitative; so frequency tables are not really helpful in exploring the data.

I am going to convert some quantitative variables into categorical variables to look at some frequency tables.

For the frequency table on alcohol consumption I have decided to split the variable in ranges of 3 (litres of pure alcohol per capita); so the lowest group includes countries with 3 litres or less, the second group between 3.01 and 6 and so forth.

Alcohol consumption frequency table:

From what we can see almost half of the countries (82 out of 167)  lies in the first two ranges (0 to 3 litres per person and 3.01 to 6 litres per person). The distribution of this variable is postively skewed; but we will see more on that with descriptive statistics.

The SAS code to convert the quantitative alcohol consumption into a categorical variable is shown below. In this way I have created 6 buckets

/*creating alcohol consumption buckets*/

[jpassionstats]

Assignment 1: Definition of the Research Project

In this project I am going to be using the Gapminder data set as I am interested in how global trends are intertwined between each other.

According to the Gapminder code book, "Gapminder is a non-profit venture promoting suistainable global development and achievement of the United Nations Millenium Goals". Gapminder includes data for all the 192 United Nations members and has information on more than 200 indicators.

Each observation in the data set corresponds to a country.

For my analysis project I decided to look at the relationship between Alcohol Consumption and Suicide rate.

I believe there is a positive correlation between the two variable but a deeper dive into the data can shed some light on the nature of the relationship. I also believe that there are other variables associated with Alcohol consumption and Suicide rate.

My research questions are:

Is there an association between alcohol consumption and the suicide rate?

If yes, is the association between alcohol consumption and suicide rate affected by other variables? In particular, I will be looking at how Employement group affect the relationship between alcohol consumption and suicide rate.

From the Gapminder code book, here the variable definition:

Alcohol consumption: 2008 alcohol consumption per adult (age 15+), litres Recorded and estimated average alcohol consumption, adult (15+) per capita consumption in litres pure alcohol.

Suicide rate: 2005 Suicide, age adjusted, per 100 000 Mortality due to self-inflicted injury, per 100 000 standard population, age adjusted.

Employement rate: 2007 total employees age 15+ (% of population) Percentage of total population, age above 15, that has been employed during the given year.

[the-path-to-procrastination]

Quiz 6 result

Week 5 Quiz was a short but difficult one: I got 8/9 on the first attempt and 9/9 on the second one.

[the-path-to-procrastination]

Quiz 5 result

Week 4 Quiz 2 of 2 was long, but easier: 18/18 on the first attempt!

I think I'm doing well...