Homework assignment QDA II. in winter 2014

Back to the main page    Quantitative Data Analysis II. (presentations, readings, etc.)
What we have done QDA II. 2014 - winter semester
Homework must be writen in doc MS Office format (possibly in txt, rtf or pdf) with your answers to the questions and brief interpretation of the results (select only adequate results from SPSS output).
Send it to the email jiri.safrATseznam.cz; in the subject, please write: QDA2, homework NO., YOUR NAME

1. homework (18/10/2014) Computation of standard error of mean and Confidence Intervals for means (by hand)
(assignment in process)
2. homework (22/10/2014) Confidence Intervals for means within subgroups (using SPSS)
Data: ISSP 2007      
Find out using computation of Confidence intervals (95 %), whether there are (statistically significant) differences in Political orientation on Left-Right scale [q37] among:
a) age groupings [Age4]
b) marital status [s10]  
Among which groups we can find differences (at p < 5%)? Make a brief sociological interpretation of the results. And when you know the differences of Political orientation along the age (as well as age according marital status - make a crosstabulation to see!), how can you (or can't) interpret differences along the marital status categories?
Use Examine/Explore command and also graphical visualization with ErrorBar. All you need can be found in Syntax 22/10/2014 (computation Confidence intervals for means in SPSS) and in the handout-presentation 3. Confidence interval, Standard error (1.)  (in pdf)  (from page ca 31).
3. homework (??/11/2014)
Data: ISSP 2007  
(this homework is joined with homework no. ?? on testing of statistical hypotheses)
4. homework (5/11/2014) Interval estimation of proportion (CI %) of citizens admitting Christian religion [s28]   Data: ISSP 2007 Procedure: First, recode variable s28 into the dichotomy of "Christians" by collapsing categories 1, 2, 3, 4 (into the new category 1 = Christians) and 5, 6 (into 0 = other).
Calculate the 95 % confidence interval for the proportion of Christians, namely:
a) by hand via inserting into formulas - instructions can be found in the presentation Standard error and confidence intervals (2.) – for proportion (slides ca 14-16)
b) and also verify the result in SPSS using syntax routine for Calculation of confidence interval of proportion/percentage For simple CI of % choose the second test: Large-Sample Confidence Interval for a Single Population Proportion (just enter: n and p).
 You can also find it (the selected test) with further comments in Syntax 19/11/2014. Describe the result, i.e. what the confidence interval means.
c) Bivariate research question: Are there differences (at 5% of statistical significance) in proportion of voters [q34] along groups of with/without Christian religious belief? Be careful, the dependent variable is here Voted/Didn't vote [q34] whereas Christian/otherwise is independent one. (A hint: You can solve it via simple bivariate 2x2 crosstabulation (CROSSTABS) and filling the proportions of voters independently within a groups of Christians/otherwise e.g. into the second syntax routine Large-Sample Confidence Interval for a Single Population Proportion). The problem can be solved by testing whether the differences between proportion in subgroups p1-p2 is zero, for this use the last-forth test Large-sample Confidence Intervals for Comparing for two population proportions).
Note: G. Pryce [2002] syntax routine doesn't work in PSSP (yet) but you can use web calculators: Confidence Interval of a Proportion    Confidence Interval for the Difference Between Two Independent Proportions  
5. homework (12/11/2014) Assessment of time change indicated in survey sample data using confidence intervals for proportion/percentage, i.e. comparison of your own results (from micro-data) with previous results already published   Data: ISSP 2007 (en) and results from EVS 1991, CR (European Values Study)
In 1991 in the Czech Republic, the EVS survey reports this participation rate, i.e. membership in:
 1. Religious/ church organization 5.4%
 2. Political parties/ movements 5.0%
 3. Sport and entertainment organization 17.8%
 4. Cultural, arts organizations 5.6%
EVS 1991 sample size was N = 2109.
Compare these results from 1991 with results from the ISSP 2007 survey, where membership - participation in organizations is represented by these variables:
Church or other religious organization [q13_c], Political party or association [q13_e], Sports club or association [q13_a] and Cultural association [q13_b]. To indicate membership consider: for "member" (=1) a sum of answers from 1 (1x a week) to 4 (1-2x a year) and for "non-membership" (=0) category 5 (not even once). You don't need to recode original variables first (but you can do it using RECODE).
Find out using computation of confidence intervals, whether and possibly at which organizations, the civic participation rate has changed in the Czech society during the last 16 years.
A hint: first for 2007 data compute descriptive tables (Frequencies), you even don't need to recode variable first (just sum corresponding categories counts). Second, compute confidential intervals to detect change, for which there are two strategies: 1. compute confidential intervals for 1991 as well as 2007 and then inspect whether they overlap; or 2. you can also compute confidence interval only for the difference (1991-2007) then you inspect wether it containes zero (if yes then we can't confirm the change).
For both strategies you can use SPSS syntax routine by G. Pryce [2002] Calculation of confidence interval of proportion/percentage  For the 1. method - simple CI of % choose the second test: Large-Sample Confidence Interval for a Single Population Proportion (just enter: n and p); for the 2. method use the forth-last test: Large-sample Confidence Intervals for Comparing for two population proportions   You can find both tests prepared with my commentaries in Syntax 26/11/2014
Note that this syntax routine doesn't work in PSSP (yet) but you can use web calculators: Confidence Interval of a Proportion    Confidence Interval for the Difference Between Two Independent Proportions  or you can compute CI by hand (it is very simple, see the presentation).
6. homework (3/1/2014) Introduction of statistical hypothesis testing - tests of means for numeric variables )    Data: ISSP 2007 (en)     
1. Read obligatorily: Testing Hypotheses (chapter 7) in Leon-Guerrero, Anna, Chava Frankfort-Nachmias. 2012. Essentials of social statistics for a diverse society. Thousand Oaks (Calif.): SAGE Publications.
2. Tests for means: using T-test and simple One-way ANOVA (F-test) find out, whether there are (statistically significant at p < 0.05) differences in Political orientation on Left-Right scale [q37]:
 a) between men and women [s30] - Are women more on the Right than men?  A hint: Use Independent-samples T-test (T-TEST GROUPS).
 a) among age groups [Age4]  A hint: Use simple analysis of variance (ONEWAY). Further, try to find out using Post-Hoc tests (e.g. Bonferroni), which age groups are statisticaly different from others. Is Right-political orientation increasing with the age?
In both cases use in addition to a statistical test graph with means of L/R orientation along the categories of the independent variables (gender, age) with 95% confidence intervals (ERRORBAR). Here it is the same problem-question as you have solved in the homework np. 2 (22/10/14).
All you need you can find in Syntax 10/12/2014 (T-tests an Oneway ANOVA)  and also in new presentation Statistical hypothesis testing (1) - principle and tests for numerical variables  (pdf version) (slides ca 29-38) and Testing Hypotheses (chapter 7) in Leon-Guerrero, Anna, Chava Frankfort-Nachmias. 2012. Essentials of social statistics for a diverse society. Thousand Oaks (Calif.): SAGE Publications..

And don't forget to state in the subject of your email: QDA2, Homework NUMBER, YOUR NAME

Back to the main page    Quantitative Data Analysis II. (presentations, readings, etc.)
What we have done QDA II. 2014 - winter semester