.GSS 2014 Statistical Analysis: Written up Assignment
Introduction
The GSS 2014 has been established with the view to generate ongoing research data on demographics of the country for social welfare and decision making. The General Social Surveys (GSS) have been conducted by the National Opinion Research Centre (NORC) annually since 1972, except for the years 1979, 1981, and 1992, and biennially beginning in 1994. The GSS surveys are supposed to be part of a social indicator research, applying questionnaire items and wording in order to facilitate time-trend studies. There are a total of 3,842 cases in the data set. The present paper draws in from the research conducted as part of GSS 2014 in terms of the data and direction. The paper poses certain research questions and drawing assumptions from the same, answers the questions based on a thorough statistical analysis of the variables involved in the study and as defined in GSS 2014.
Methods
The present paper analyses the data collected as part of the ongoing General Social Survey and specifically deals with the data pertaining to 2014. The topics covered as part of the demographic research necessitate a variety of analytical techniques. The sample size for this research is 3842. The research area is the United States of America. The data provided is primarily frequency data for the various variables as defined in the survey which include both categorical and interval data. The categorical variables and their levels have been coded in numerical form for analysis purposes. The paper derives various research questions and assumptions as part of the analysis process as mentioned below. The various research questions are means to gain insights into the demographic interrelationships of various factors .
Research Questions
Does regional background have a relationship with the education level of a person?
Does parental education level predict a person’s work situation in terms of Income and Class?
Does parent immigration status predict a person’s work situation in terms of Income and Class?
Does political orientation predict voting behaviour?
Does political orientation of a person predict a person’s perception of people or life?
Does income level of a population help predict the general health of the population?
Does regional background of a person affect his or her opinion on various issues?
Does income of a person affect the job satisfaction of the person?
Hypotheses
H1: Regional background has a significant relationship with the education level of a person.
H2: The education level of the parents of a person significantly affects a person’s income and class.
H3: The immigration status of a person’s parents has a significant effect on the person’s work situation in terms of the income and class.
H4: The political orientation of a person has a significant relationship with the voting behaviour of a person.
H5: The political orientation of a person is a significant predictor of a perception of people and life.
H6: The income level of a population has a significant positive relationship with the general health of the population.
H7: The regional background of a person significantly affects his or her opinion on various issues.
H8: The income of a person affects the job satisfaction of a person in a significant manner.
The hypotheses can further be dissected into null and alternative hypotheses statements as mentioned below.
Ho1: Regional background does not have a significant relationship with the education level of a person.
Ha1: Regional background has a significant relationship with the education level of a person.
Ho2: The education level of the parents of a person does not significantly affect a person’s income and class.
Ha2: The education level of the parents of a person significantly affects a person’s income and class.
Ho3: The immigration status of a person’s parents does not have a significant effect on the person’s work situation in terms of the income and class.
Ha3: The immigration status of a person’s parents has a significant effect on the person’s work situation in terms of the income and class.
Ho4: The political orientation of a person does not have a significant relationship with the voting behaviour of a person.
Ha4: The political orientation of a person has a significant relationship with the voting behaviour of a person.
Ho5: The political orientation of a person is not a significant predictor of a perception of people and life.
Ha5: The political orientation of a person is a significant predictor of a perception of people and life.
Ho6: The income level of a population does not have a significant positive relationship with the general health of the population.
Ha6: The income level of a population has a significant positive relationship with the general health of the population.
Ho7: The regional background of a person does not significantly affect his or her opinion on various issues.
Ha7: The regional background of a person significantly affects his or her opinion on various issues.
Ho8: The income level of a person does not affect job satisfaction of a person in a significant manner.
Ha8: The income level of a person affects job satisfaction of a person in a significant manner.
Analysis & Discussion
The overall composition of the population based on the gender of the people is as shown below.
The females slightly outnumber males in the survey is a healthy sign of a balanced population.
The average age of the sample population is as shown below.
An average middle age of the respondents with a relatively low standard deviation shows that the population is pretty mature and well disposed to respond to the queries in the survey.
The regional composition of the sample population is as shown also pretty representative of the actual population and the various characteristics of the regional composition are as shown below .
As shown above the South Atlantic region has the greatest representation of the population, while New England has the lowest.
As far as the educational status of the population is concerned the highest degree earned is tabulated as below.
As the above table shows, the graduate respondents are a smallest percentage of the overall population while the high school seems to be the highest degree of the population apparently. Thus the education is not at the desired levels going by numbers.
A comparative view of the sample populations based on the regional background and education levels is as shown below.
The above table shows that the regional background of the population in terms of the frequencies under major levels or groups is more evenly closely and evenly distributed around the mean (SD=212) than the education level (SD=663).
Further, we proceed to testing the hypotheses as propounded as part of the paper and as mentioned above.
Hypothesis H1
The variables REGION and DEGREE were tested for comparison of the means, as the range of values have been numerically defined and REGION is treated here as an independent interval variable, which has a relationship with the level of education denoted by the variable DEGREE. An independent samples t-test was conducted assuming unequal variances and the results are tabulated as below.
As is evident from the above table, there is statistically significant difference (t=77,p=0) thus the null hypothesis Ho1 is rejected and we conclude that regional background has a significant relationship with the education levels.
Hypothesis H2
Moving further on, the Parental education levels were studied as predictors of person’s work situation. Specifically, the variables PADEG and MADEG were combined into a single variable called PRADEG by calculating the average of the scores on the former variables, since both had similar number and nature of levels. Further on a Chi-Square test was conducted on PRADEG and CLASS, which denotes the social status of the respondent. The results are tabulated as shown below.
Chi-Square Test for Two sample for Means
As observed above there is no statistically significant relationship between the educational levels of parents and the social class and status. Further, the educational levels of parents were examined against the overall income levels. Thus a two independent samples t-test was conducted on the variables PRADEG and RINCOME. The results are as tabulated below.
Two samples T-Test
As is evident from the above table, there is no statistically significant difference (t= -35.677, p=6.5E-243) thus we conclude that PRADEG and RINCOME have a significant relationship with each other. Based on the results of above two tests, we can infer that the education level of the parents of a person significantly affects a person’s income and class.
Hypothesis H3
Moving further on, we tested more hypotheses as moved on. This time the parental education levels were substituted by the immigration status of the parents of the respondent. Specifically, the variable as defined in the GSS as PARBORN was compared with the variables CLASS and RINCOME. PARBORN being a categorical independent variable and CLASS being a categorical dependent variable, a chi square test was conducted on the frequencies of the two variables. The results are tabulated as below.
Chi-Square Test
As is evident from the above test, the relationship between the variables PARBORN and CLASS is not statistically significant (Chi-Square=-.06, p=0.28) at 0.05 significance level. Thus the immigration status of the parents does not have a significant relation with the social class or status of the respondents.
Further, the variable RINCOME was examined against the variable PARBORN to measure the effect of parents’ immigration status on the collective income of the respondents. A one way ANOVA was conducted to measure the difference in the mean values of the two variables. The results were tabulated as shown below.
As shown above the variables have a huge difference in means, although the mean difference is not absolute, thus the variance has to be considered as an appropriate measure of comparison. There is a huge difference in the variances as well. As also evident from the test there is a statistically significant difference between the mean values of the two variables PARBORN and RINCOME. Thus, the null hypothesis Ho3 can not be rejected and thus the immigration status of a person’s parents does not have a significant effect on the person’s work situation in terms of the income and class.
Hypothesis H4
Further, the paper attempted to determine the effect of the political orientation of a person on the voting behaviour. Specifically, the voting behaviour in the 2008 and 2012 polls was compared with the party voted for and political views measured as PARTYID and POLVIEWS. The voting behaviour in the two polls was coded as VOTE08 and VOTE12. The two variables were combined into one variable named VOTE with three levels as shown below.
Did not Vote in Either-1
Voted in both-2
Voted in one-3
A chi-square test was carried out on the variables PARTYID and VOTE and the results presented as below.
As shown above, the variance between the two variables is not considerable. This is corroborated by the significance test (Chi-square=31, p=0.00), which shows that the relationship between the two variables is statistically significant at 95% confidence level.
Further, a factorial two way ANOVA was administered on the variables POLVIEWS and VOTE and the results are presented as below.
Based on the above table, there is strong relationship between the two variables as the variances do not differ considerably and this is supported by a high F value (F=0.95).
Thus the null hypothesis Ho4: The political orientation of a person does not have a significant relationship with the voting behaviour of a person-is rejected and it is inferred that the political orientation of a person from sample population has a significant relationship with the voting behaviour.
Hypothesis H5
The paper moves on to measure the effect of the political orientation on the perception of the respondents about people and life in general. Specifically, the relationship between the variables POLVIEWS and TRUST was examined. A Chi square test was conducted between POLVIEWS and TRUST. The results were tabulated as below.
As shown above, the difference in variance between the values of POLVIEWS and TRUST is not considerable and the same is supported by the high Chi-Square value significant at .05 % significance level (p=0.00).
Based on the above results, the null hypothesis Ho5: The political orientation of a person is not a significant predictor of a perception of people and life - is rejected. Thus, the political orientation of a person does seem to have a significant prediction value for the respondent’s perception about people and life.
Hypothesis H6
Moving ahead the paper explores the effect of income levels on health. Specifically, the variable INCOME as defined in thee GSS 2014, was examined against the variable HEALTH. A Pearson’s coefficient of correlation was calculated between the two variables as shown below.
Correlation
r=.02
We can see that HEALTH and INCOME share just 4% (r^2*100) of their variability. Thus there is no strong relationship between the health and income of the respondent population. Therefore the null hypothesis Ho6: The income level of a population does not have a significant positive relationship with the general health of the population – can not be rejected. Thus it is inferred that health and income levels do not share a positive relationship for the sample population.
Hypothesis H7
Moving further, we examine the effect of region of country on the various opinion related questions. In other words we set out to find whether the regional background of a person influences his or her opinions. Specifically, the variable REGION was examined against the variables FEPOL measuring the view that women are not suited for politics and SEXEDUC measuring the opinion on sex education in the country. The variable REGION was reduced from 9 levels to more manageable 5 levels and renamed as REGION1. A two way factorial ANOVA was conducted on REGION1, FEPOL and SEXEDUC. The results were tabulated as shown below.
Two way factorial ANOVA
The above results indicate that an interaction between the variables REGION1 FEPOL and SEXEDUC yields a significant relationship effect (F=2.84, p=6.7E-141).
Thus we reject the null hypothesis Ho7: The regional background of a person does not significantly affect his or her opinion on various issues- is rejected. Thus it is inferred that the regional background of a person does affect the opinion of a person on social issues.
Hypothesis H8
Finally, we move to the last part of the analysis wherein we studied the effect of income level on the job satisfaction of a respondent. Specifically the variables JOBSAT pertaining to satisfaction in work done and JOBSAT1 pertaining to satisfaction in overall job were examined against variable INCOME. A two way factorial ANOVA was administered. The results were tabulated as hereunder.
Factorial (Two Way) ANOVA
The above results indicate that overall model is statistically significant (F=1.76, p=5.39E-70). Thus the interaction between the variables yields a statistically significant relationship. Therefore, we reject the null hypothesis Ho8: The income level of a person does not affect job satisfaction of a person in a significant manner. Thus it is inferred that the income level of a person does affect the job satisfaction of the person in a significant manner.
Conclusion
Finally speaking, the various research questions have been answered and the results drawn. Specifically, we conclude that regional background has a significant relationship with the education levels. The education level of the parents of a person significantly affects a person’s income and class. As far as the immigration status of a person’s parents is concerned it does not have a significant effect on the person’s work situation in terms of the income and class. The political orientation of a person from sample population does have a significant relationship with the voting behaviour. The political orientation of a person does seem to have a significant prediction value for the respondent’s perception about people and life. However, health and income levels do not share a positive relationship for the sample population. Further, the regional background of a person does affect the opinion of a person on social issues. Finally, the income level of a person does affect the job satisfaction of the person in a significant manner.
References
ARDA. (2014). General Social Survey 2014 Cross-Section and Panel Combined. Retrieved from http://www.thearda.com/: http://www.thearda.com/Archive/Files/Codebooks/GSS2014_CB.asp
Armstrong, R., Eperjesi, F., & Gilmartin, B. (2002). The Applicattion of analysis of variance (ANOVA) to dfferent experimental designs in optometry. Opthal.Physiol., 248-256.
Cohen, B. H. (2002). Calculating a Factorial ANOVA from means and Standard Deviations. Retrieved from http://www.psych.nyu.edu/: http://www.psych.nyu.edu/cohen/Calc_ANOVA.pdf
IDRE. (2016). What statistical analysis should I use? Retrieved from http://www.ats.ucla.edu: http://www.ats.ucla.edu/stat/spss/whatstat/
West, M. D. (2008). Use of the Chi-Square Staistic. Retrieved from http://ocw.jhsph.edu/: http://ocw.jhsph.edu/courses/fundepiii/pdfs/lecture17.pdf
Zhou, K. H., Tuncali, K., & Silverman, S. G. (2003). Correlation and Simple Linear Regression. Radiology, 617-628.