Reference no: EM132342267
Assignment - The assignment should be done using R studio and should be sent to be solved in .rmd format.
Instructions: For Project, you will use your linear regression tools to analyze data and build a report. To do this, you will generate a full data analysis workflow from articulating a research question to loading the data to exploring and visualizing your variables to building a model with diagnostics and possible transformations to interpreting your final results. Think of it as a data analysis story, like a short research paper you might submit to a scientific journal for publication, or like a data report you might submit to a client you're consulting for.
Your data analysis will be written up in R-Markdown on the departmental shimmer server, similarly to how you have been doing homework's and labs, but your project should be structured and presented like a report you would submit to a client or an academic journal (with full sentences and with written flow).
Some sections (Introduction, Discussion) only require written text, while others (EDA, Modeling, Prediction) will be a combination of code chunks and written text. All the code you need can be found in your homework's and labs and will just require editing; there is no need to write brand new code.
An example project is posted. There is a template markdown file you can download to your computer and then upload and open in R-Studio on the shimmer server. It is recommended that as you start working, you re-name the file something else (like maybe 202proj1ver1.Rmd or something), to avoid accidentally over-writing your project with the original blank template. Also, as you work, it is similarly a good idea to re-name the versions after a major revision. Each student will select from one of four project topics. Each project topic has a research scenario with variable description and a data file. In each research scenario, there are specific research questions/hypotheses that identify the response variable and the three predictor variables.
You will build a multiple regression model using the specified predictors to model the specified response.
Note: These are real datasets, and hence your regression diagnostics/p-values/R2 values/etcprobably won't be very good. The goal isn't a 'perfect' model or a very significant p-value, the goal is to explain your choices, demonstrate understanding of the techniques, and produce as reasonable a model you can (within the constraints of the assignment and your skills up to this point).
THE TOPIC IS: Court Cases: Predicting the amount awarded to the plaintiff.
Your report should have the following 5-section structure:
- Introduction: Motivate why the topic is interesting or important, and what your overall research question is. The introduction should not have methodological discussion, it should just be the place to catch the reader's interest. (This is a brief section that only requires text. Use complete sentences).
- Exploratory Data Analysis: Describe your data set. Indicate where the data came from and how it was collected (if known), as well as what the variables are (with units); Use written text along with numerical summaries and graphs. Perform univariate and bivariate EDA. Remember that your EDA should support your modeling section and results.
- Modeling: Use your linear regression tools to build a model predicting your response variable. Use diagnostics to check your model assumptions; use variable transformations if needed and appropriate/justifiable. Remember to properly justify your choices and actions. Don't show the output for all the models you try, only focus on your final chosen model and justify why you ended up choosing that particular model (what other models did you try, and why did you end up selecting the final one as most appropriate?). This section should be a combination of code, diagnostic graphs, output, and written text.
- Prediction: Calculating specific prediction specified in each scenario. Note: If you include transformations in your model, then to keep things simple you can compute and interpret the prediction in terms of the transformed variables. Also for simplicity, you can compute the prediction "by hand" (in a code chunk, using R like a calculator) just plugging in the values. Note also, you might decide that your best model doesn't use all the explanatory variables, in which case specified values of those variables won't play a role in your prediction. (This section should be a combination of code, output, and written text.)
- Discussion: Short section where you give your overall conclusionsin terms of the problem of interest. Also, it is important to critique your own analysis - what limitations does your model have, and why? (Issues with diagnostics that should be especially pointed out? Issues with the data that should be mentioned? Other issues that should be mentioned?) Finally, projects should briefly discuss future directions the work can go. (What else would you have liked to do?) (This is a brief section that only requires text. Use complete sentences).
Attachment:- Statistics Assignment Files.rar