Data analysis report

Trust in authorities

Risk perception

Methodology

Measurement

A survey was conducted to undergraduate students of City University of Hong Kong. The questionnaire designed by investigator Jacky Chau-kiu Cheung and was designed to inquire about the students’ perception of the risk, attitudes towards the financial crisis, SARS-like epidemic and terrorist, government trust and so on. Other variables such as risk of SARS-like epidemic and risk of terrorist attack were also included in the questionnaire. Totally the questionnaire has 90 items contained the items on the perceptions and attitudes about risks and the main background information.

Sample Size

The questionnaire was administered in class between **and ** and resulted in a sample size of 350 undergraduates. Majority of the respondents were students from the main programs of the Department of Applied Social Sciences were surveyed. They were primarily year 1, 2, 3 and 4 students enrolled in the criminology, psychology, counseling, sociology and social work programs. The education levels of participants are range from the bachelor's level, master's level and doctoral level.

Data analysis

Input data using the Statistical software.

For points:

1. Factors analysis used to select the impact factors to the risk perception.

2. Normality test of all the demographic variables.

3. New components were computed after the “Factor analysis”, then test new normality.

4. ANOVA used to test the difference in means among the different major groups.

5. Correlation would be tested among all the variables in this research.

1) Independent variables: “Knowledge”, “Detectability, “Controllability of government”, “Trust in authorities”, “demographic characteristics”.

2) Dependent variable “perceived risk of financial crisis”, “SARS-like epidemic”, and “terrorist”.

6. Univariate linear regression used in the three dependent variables.

7. Multivariate linear regression model used in outcome variables.

8. Logistic regression model used in the three DVS and use the IV as predictors respectively.

Result

Factor Analysis: Reliability

IV: Knowledge

From the two table, the correlation of the three items are not obvious/correlated based on the Cronbach's Alpha is .478, and one item deleted Alpha is .397, .325, and .4.19. All of them are not at the significant correlation level >.60. Based on the factor analysis result, I will extract the financial result as my focus and specific my variables to “Knowledge about financial crisis”. Also for this reason, my other independent variables “Detectability” and “Controllability of government” also will be materialized to “Detectability about financial crisis” and “Controllability of government about financial crisis”. Moreover, will the total score of selected items as an index/scale score.

IV: Distrust in authorities

From the two tables, the correlation of the three items is well correlated based on the Cronbach's Alpha is .761, and one item deleted Alpha is .625, .602, and .787. All of them are at the significant correlation level >.60. That means I can use the three items total score to measure the independent variable “Distrust of authorities”.

IV: Trust in authorities

From the two tables, the correlation of the three items is well correlated based on the Cronbach's Alpha is .693which is at the significant correlation level >.60, and one item deleted Alpha is .573, .600, and .624. That means I can use the three items total score to measure the independent variable “Trust in authorities”.

Distribution

Frequencies

From this chart and based on the sample is 394 and valid sample size is 382 which means the sample is large enough besides that the independent variables, demographic variables, and dependent variables retained a fair skewness (<3) and kurtosis (<10) level. It is standard normal distribution curve.

T-Test

ANOVA

From this table, it showed only the level has a significant difference between groups based on the one way ANOVA test P value is .002 which is <.05, while the major and age do not show the significant difference between groups in perceived risk of financial crisis to self.

Correlations between “Knowledge about financial crisis”, “The detectability of financial crisis”, “The controllability of financial crisis by government”, “The risk of financial crisis to society”, “The risk of financial crisis to you”, ‘Distrust in authorities”, and “Trust in authorities” were analyzed as result showed in Table.

“Knowledge about financial crisis”, “The detectability of financial crisis”, “The controllability of financial crisis by government”, and ‘Distrust in authorities” were positively correlated with perceived financial crisis risk to society (p < .01). While the “The risk of financial crisis to society’ and the “Trust in authorities” do not show a significant correlation with each other due to the P value is .892 which is much higher than the (p < .01). Besides, “The risk of financial crisis to self” and the “Distrust in authorities” also do not show a significant correlation with each other due to the P value is .072which is a little higher than the (p < .05).

“Knowledge about financial crisis” and “The detectability of financial crisis” showed a high correlation with each other at the Pearson Correlation coefficient is .440. Meanwhile, the two dependent variables also showed the high level of correlation with each other at the Pearson Correlation coefficient is .422.

I. “Knowledge about financial crisis” has correlation with “The risk of financial crisis to society” and “The risk of financial crisis to you” respectively is .147** and .247** at a significant different level .001.

II. “The detectability of financial crisis” has correlation with “The risk of financial crisis to society” and “The risk of financial crisis to you” respectively is .224** and .246** at a significant different level .001.

“The controllability of financial crisis by government” has correlation with “The risk of financial crisis to society” and “The risk of financial crisis to you” respectively is .144** and .118* at a significant different level .001 and .05 respectively.

Regression 1

In predicting the perceived risk of financial crisis to society using a univariate linear regression analysis, significant predictors are “The detectability of financial crisis” .004 (p < .01), ‘Distrust in authorities”.002 (p < .01) and “Major” .006(p < .01). All the significant predictors exhibited positive effects on perceived risk of financial crisis to society. The strongest effect of “perceived risk of financial crisis to society” is the independent variable “Major of study” (Beta= .355).

Regression 2

In predicting the “perceived risk of financial crisis to self” using a univariate linear regression analysis, significant predictors are “Knowledge about financial crisis” .003 (p < .01) and ‘The detectability of financial crisis”.003 (p < .01). Both of the significant predictors exhibited positive effects on the “perceived risk of financial crisis to self”, while both of them not show a strong effect of “perceived risk of financial crisis to self” based on the Beta is equal to .181and .177 respectively.

Discussion

Based on the SPSS analysis result, the following conclusions emerge.

First, “Knowledge about financial crisis” has a positive correlation with “perceived risk of financial crisis to self”. Which also can support the H1 is correct (H1: The higher level of risk knowledge, the lower the risk perception will be.) However, “Knowledge about financial crisis” has not showed the significant correlation with the “perceived risk of financial crisis to society”. It also means when the risk represented the risk to self, the H1 while be proved, while, when the risk represent the risk to society, the H1 would not be proved.

Second, “The detectability of financial crisis” has a positive correlation with “perceived risk of financial crisis to self” and “perceived risk of financial crisis to society”. Which also can support the H2 is correct (H2: The more risk detectability, the higher risk perception will be.)

Third, “The controllability of risk by government” has not showed a significant correlation with the “perceived risk of financial crisis to self” and “perceived risk of financial crisis to society”. It also means H3 cannot support by our data and the relationship between “The controllability of risk by government” and “risk perception” is not clear.

Fourth, “Distrust of the authorities” has a positive correlation with “perceived risk of financial crisis to society”. While the “Trust of the authorities” has not showed a significant correlation with “perceived risk of financial crisis to society”. Neither “distrust of the authorities” nor “Trust of the authorities” showed a significant correlation with the “perceived risk of financial crisis to self”. It also means the H4 cannot be proved based on our limited data. (H4: The more trust of the authorities, the less risk perception will be.)

Fifth: At the demographic aspect, only “Major of study” showed a significant correlation with the “perceived risk of financial crisis to society”. But not showed a significant correlation with the “perceived risk of financial crisis to self”. Besides that other demographic aspect do not showed the significant difference with two dependent variables. It also indicates that the major of study can impact the risk perception.

Reference

Roger E. Kasperson,' Ortwin Renn,' Paul Slovic,2 Halina S. Brown,' Jacque Emel,'Robert Goble,' Jeanne X. Kasperson,'~a~n d Samuel Ratick'. 1987.The Social Amplification of Risk A Conceptual Framework. Revised January 8, I988

Pilarski, R. (2009). Risk perception among women at risk for hereditary breast and ovarian cancer. Journal of Genetic Counseling, 18(4), 303–312.

Wibecke Brun, Cognitive components in risk perception: natural versus man made risks. Journal of Behavioral Decision Making, 1992,5: 117-132

Wiersema,M ,Bantel,K．Top management team demography and corporate strategic change[J]．Academy of Management Journal,1992,35(1):91－121

Norusis, Maruha J. 2005. SPSS 14.0 Advanced Statistical Procedures Companion. Upper Saddle River, NJ: Prentice Hall.

Hawkins, Daniel N., Paul R. Amato, and Valarie King. 2006. “Parent Adolescent Involvement: The Relative Influence of Parent Gender and Residence.” Journal of Marriage & Family 68(1):125-136

1. Chi-Square Test

a. 1 cells (25.0%) have expected frequencies less than 5.

b. The minimum expected cell frequency is 3.8.

A chi-square test of the major field of study P=0.008 indicated that the four fields were not equally likely among the students in a statistical sense which is (p < .05) significant different from the expected distribution. Accordingly, the chance of being a sociology student has a greatest chance and the chance of being a counseling student was least. This finding may just describe the population distribution.

2. Reliability

The chart showed the three independent variables of a17, a27, and a37 are satisfactory internal consistency because the Cronbach's Alpha is 0.654, but from the second chart, by excluding the items a17, a27, or a37 separately the Cronbach's Alpha changed to (.572, .489, or .628) which are <.654, that indicated that the reliability cannot be improved by excluding the items in this data set.

3. Factor analysis

A factor analysis about eight items about financial risk suggested two factors. The two-factor solution represented a good fit to the data, based on a chi-square test, using the maximum likelihood extraction method (p = .524).

The factor analysis using the varimax with Kaiser Normalization rotation method and using a rotation converged in 3 iterations, the first factor consisted of (a29), (a30), (a31), and (a32); the second factor undergirded (a33), (a34), (a35), and (a36). The perception about (a31) the personal responsibility for terrorist attack was the most important item to reflect the first factor. It appears that the first factor is about the prominence of risk.

4: Regression

Part1:

It shows Regression Standardized Residual is decreasing variation with age, is a sign of heteroskedasticity.

It shows Regression Standardized Residual Variation with Control roughly equal across controllability levels, it supporting the homoscedasticity.

Part 2:

It is repeating the previous finding of heteroskedasticity, using the residualized age, which is independent of controllability to the dependent variable of risk.

compute risk.r=age*risk.r.

graph scatterplot= age with risk.r .

It shows after weighting the residual by age, it would produce homoskedasticity.

temporary.

compute both=mean(control,new).

regression /missing mean /statistics default

/dependent risk /enter both new age female.

It showing the Regression standardized residual of the dependent variable risk has a normal distribution.

The independent variable of new, age, and female has a R square is (.195) which can have a strong confident to use the three predictors to predicts the dependent variable of risk. From the chart, the independent variable of new, age, and female (p<.05) has a significant effect on the dependent variable of risk.

The independent variables “Both” and “New” have significant effects on the dependent variable “Risk”. The regression coefficient of “Both” is .361 (p=.001) which indicates that for every one unit increase in “Both” will result in .361 units in “Risk”. Whereas, the regression coefficient of “New” is .214 (p=.018) which means that for every one unit increase in “New” will result in .214 units in “Risk”. The independent variables “Age” and “Female” do not have significant effects (P > .05) on the dependent variable “Risk”. It also means that “Age” and “Female” are not significant predictors of “Risk”.

Part 4:

Based on the first chart it indicates that the independent variables of “Three”, “New”, “Age”, and “Female” have R square equal to .196 which can have a strong confident to use the four predictors to predicts the dependent variable of risk.

The second chart it indicates the independent variable of “Three”, “New”, “Age” (p<.05) has a significant effect on the dependent variable of risk, while the independent variable “Female” do not have significant effects (P > .05) on the dependent variable “Risk.

However, Comprehensive front three charts, the independent variables “Three”, “New”, “Age”, have significant effects on the dependent variable “Risk”. The regression coefficient of Three”, is .751 (p=.001) which indicates that for every one unit increase in “Three” will result in .751 increase units in “Risk”. Whereas, the regression coefficient of “Age” is -.044 (p=.023) which means that for every one unit increase in “Age” will result in decrease .044 units in “Risk”.

The independent variable “Female” do not has significant effects (P > .05) on the dependent variable “Risk”. It also means that “Female” is not a significant predictor of “Risk”.

5. General Linear Model

temporary.

compute inter=zcontrol*znew.

regression /missing mean /statistics default tol

/dependent risk /enter control new age female inter

The general linear model that added the interaction between the control and new as a predictor turned out that this interaction did not generally make a significant difference (p>0.05) in the dual outcomes of perceived risks to self and to society. Nevertheless, two particular significant interactions were that involved, the controllability of terrorist attack by government, and the perceived novelty that indicated significantly perceived risk to society.

6. Two-stage Least Squares Analysis

2sls risk with control new /control with risk /new with yr control /instrument age female yr.

In a nonrecursive, reciprocal-effect model, perceived novelty and perceived risks were causes and outcomes reciprocally, with the year of study, age, and female used as instruments to enable the two-stage least squares estimation. Regression analysis with two-stage least squares estimation showed both the reciprocal effects were significant. Accordingly, the effect of perceived novelty on perceived risk was very strong (β = .771), and the effect of perceived risk on perceived novelty was very strong too (β = .995). However, the Equation 3 is not significant.

7: General Linear Model

glm risk.soc risk by level major with age female

/print parameter opower etasq homogeneity

/plot=profile(level major)

/emmean table(level) compare adj(bonferroni)

/emmean table(major) compare adj(bonferroni)

/design level major age female.

In multivariate analysis of variance, perceived risk to oneself and perceived risk to society were outcomes predicted by the level of study, major, age, and female. Multivariate tests showed that only the major made a significance difference in the two outcomes, with an observed statistical power of .997. This power indicated that the risk of the error of wrongly acceptance of the null hypothesis of no difference was close to zero (Type I error = .003). At the same time, the effect size in terms of partial eta (η) was .048. In addition, univariate tests revealed that the major made a significant difference in both perceived risks to society and to self. The observed statistical power was .996 and .904 respectively.

Compared with the major of criminology [major=5], the social work and psychology majors showed marginally significantly lower perceived risk. Moreover, the multivariate test indicated that the effects of all predictors on all outcomes were not significant as a whole, for example, with a psychology major Type II error of 5.407 and Type I error of .302.

Based on estimated marginal means

*. The mean difference is significant at the .050 level.

b. Adjustment for multiple comparisons: Bonferroni.

Pairwise comparison of adjusted means with Bonferroni adjustment showed that the perceived risk to society was significantly lower in the sociology student than in the social work student (mean difference = -.990*, p = .000). Similarly, the perceived risk to society was significantly lower in the sociology student than in the psychology student (mean difference = -1.104*, p = .040). Besides that, Pairwise comparison of adjusted means with Bonferroni adjustment showed that the perceived risk to oneself was significantly lower in the sociology student than in the social work student (mean difference = -.866*, p = .001).

Part 3: Class participation report

In semester B, the course of Applied Social Statistical Analysis, taught me how to use the statistical reasoning to understand / analysis the social research. Besides that I also controlled the skills of using SPSS to analysis the basic statistical methods and explain the findings. This course is the most meaningful course to me, because I am a nurse in hospital and in my future work, I should use different kinds of statistic method to analysis the medical data on patients, medical staff, and medicine experiment data. I attend every class of Applied Social Statistical Analysis on time, not only because of its meaningful to be but also this course is a little hard for me to understand. I did not have any statistic foundation expect the last semester leaning of statistic learning, but thanks to Dr Jacky, he gave us lots of recommend on the course related readings and Chinese book on SPSS analysis, I had a clear idea of this course and the know how to choose the appropriate statistical methods to do the statistical analysis on sociological practices.

Though this course, at the beginning of this course I learned the data grooming method to simplifying statistical data, data cleaning method to remove the illogical and inconsistent values, data transformation method to Skew correction, recoding data, imputation missing value and weighting in order to make the data more reliability and validity. On the second phase, I mastered the knowledge of statistic testing, such as, parametric testing and nonparametric testing. Fellow that, the date reduction method followed, it contains factor analysis which contains varimax rotated Variance, exploring underlying factors and Construct validation and reliability assessment, what’s more, all the factors based on the theoretical grounds, interpretability, identified y by alternative sets of indicators, high explaining power and the distinct loadings. The clustering method to do data reduction is also important, clustering method can classifying /group variables in order to simplifying data besides that it also can identify new variables that can represent the configuration/combination variables. Based on the data handling method, I learned the data analysis method that is linear modeling. Linear modeling contains the linear regression, multivariate analysis, Path analysis through linear regression analysis and MANOVA. Regression analysis is a statistical method to determine interdependent relationships between two or more quantitative variables. It has a very wide range, according to the number of independent variables the regression analysis can be involved into regression and multiple regression analysis. Besides that according to the number of dependent variables, it can be divided into a regression analysis and multiple regression analysis; according to the relationship between the independent variables and the dependent variable, it can be divided into linear regression and nonlinear regression analysis. This section is the most important and difficult part in this course, while is a comprehensive statistical data analysis method assemble. Follow on the nonlinear modeling was taught and it is a very useful and productive statistical method. When I start doing the SPSS individual, some excited emotion came to my mind, the learning efforts of Applied Social Statistical Analysis and results obtained are obvious. I can use the data to analysis my variables and my analysis the findings.

Review

Correspondence between reviewed literature and hypotheses needs to be clearer.

Analysis

Elaborate interpretation must be present. Sociological insight is preferable.

Discussion

It needs to be lengthier, support by literature review.

Overall

Interpretation of findings is utterly necessary.

Exercise

It can point out learning gained from the exercise. Literature review is tremendously helpful.

Performance

Learning needs to be deeper, concrete, and pertinent to sociology. It can be more focused.