Программисты — удалённая работа в Москве

Data Science. Выполнить проект по анализу данных в R. The project is devoted to the study of predictors of satisfaction with the financial situation (Q50- How satisfied are you with the financial situation of your household?) on WVS data (wave 7: https://www.worldvaludvalusurvey.org/wvsdocumentat ionwv7.jsp) You yourself formulate a research question and choose the countries (no less than 20, you might take all countries available in the data) and the predictors that will help you to answer the RQ. Depending on your framework, these can be different variables. Do not forget to add control variables (like gender). You are not obliged to read the literature for this project, but if you look at some basic work in order to formulate a more meaningful question, it will be excellent. Conceptually, the work should include: Introduction. Description of the problem and research question.Hypotheses.Methodology: Describe the data, selected variables.Analysis.Conceptual interpretation of the results of the analysis. Have your hypotheses confirmed? How will you answer your question?If desired, you can also include a section with reflection on the methods: what is useful, what is difficult to use, etc. Technical part (the most important for us) of the project includes the following steps: Prepare the data. Choose the predictors for the 1st and the 2nd level. Include no less than 3 variables for the 2nd level (you might create them as aggregated versions of the individual level variables - mean income - or find some indeces for the countries, like GDP, inequality indeces, etc.) - 2 points Correctly determine the types of variables. Evaluate and describe the distributions (not from the point of view of (not) normality, but in a meaningful way: what does the form of distribution tell us about the variable?). Are all the categories saturated with data? Recode the variables if you see the need for this.2. Evaluate the share of missings. Describe them; what type of missingness is this, you think? Impute missings with any suitable method. - 1 point Clue: it's better to impute missigns for each country separately (it's faster and more accurate)3. Conduct simple bivariate tests of the outcome with the predictors you selected. Describe the results (not only significance, but also strength and direction of the relationship). Visualize where possible. - 1 point 4. Build your multilevel regression. Choose the appropriate type of the regression for your outcome. - 2 points First, check if adding the second level is justified (ICC).Then, start the model-building process. Use the Forward Selection strategy, adding variables to the model one at one. I recommend to first add all the variables of the 1st level, then start adding the second-level predictors. Do not forget to add the control variables.Choose the best model (comparing by anova or AIC). Interpret the final model technically: describe all the coefficients; do not forget to transform the coefficients is you are using a logistic model (2 балла) 5. Model Diagnostics. How well does the model fit? Check for the main problems (library performance can help a lot). Change the model if neccessary. NB: As this is a final project, your goal is to really get a good model, not to simply describe how bad your model is. - 1 point 6. Random effects. Check the random effects for the variables of the first level. If you find such an effect, add a cross-level interaction (1st x 2nd level preditors). If not – add an interaction between the 1st level predictors only. Plot it. Interpret the results - both random effects and interactions. - 2 points 7. Add appropriate plots for all the steps of your analysis. - 1 point Submit the paper in doc/pdf/html format; code attached or embedded in the text. Raw Rmd files not accepted. After getting the grade, you will have an opportunity to correct you paper and get +1 point.