MA684留学生代写、Java/Python代写代写、代做c/c++编程语言、代写programming

日期：2019-05-07 11:12

MA684 Final Project

Spring 2019

This is an individual project—please do your own work. Some discussion with other students around

computer work for the project is permitted, but you should formulate and perform the analyses on

your own, and write up your results on your own. (A good rule of thumb: only write answers when

you are completely alone.) Questions about the content of the project or programming issues can be

directed to the instructor or TF.

This project makes up a substantial proportion of your grade, so please provide an organized,

professional, and well-edited write-up. Please write up your results in paragraph form—do not simply

annotate computer output. This is a statistics class, so please present appropriate statistical detail—

identify the statistical methods that you use, explain how you reach your conclusions, report test

statistics along with P-values to make it clear what information is being reported. Please report your

results in the context of the problem. If your write-up is complete, we should not need to refer to your

computer output, but please submit your computer output as an appendix with the project. (The

syntax will be submitted on BlackBoard, while the report will be turned in on paper at the final exam.)

You will be graded on (1) running the analyses correctly (worth 50%), (2) providing correct syntax

(all analyses must be run using syntax, but for small calculations such as p-values or effect sizes; worth

10%), and (3) your professional write-up of the solutions (worth 40%). The final write-up should be

written as though it were a final report being provided to a research client…ideally, your write-up will

be complete enough so we won’t have to refer to the computer output. The document should follow

the 5 C’s of communication: clear, concise, correct, cogent & comprehensive.

The project is due Thursday May 9 (3:00pm). Projects should be submitted at start of final exam

(printout report) and submit syntax on BlackBoard. If you encounter any problems, please email a

copy of your final project to both the instructor (harbaugh@bu.edu) and the TF

and please include your name on the project.

You are running analyses for a moderately-sized company located in the Pacific Northwest. (Note, this

is a hypothetical scenario with hypothetical data.) The company states that it values diversity and

equal opportunities for all of its employees, regardless of gender, race or any other categorization.

They have hired you to help them examine their employee evaluation and promotion process.

The employees in this data set were with the company for at least two years (one full year beyond the

probationary period), and this subset of the data includes only employees classified at the supervisory

level or below (no managers or executives) at the time of the most recent evaluation cycle. The data

on each employee is the following:

 emp_ID: a unique code for each employee in the data set

 jobrating: a score assigned by the employee’s direct supervisor on a 0–100 scale.

 salary: 12-month adjusted FT salary, in USD

 gender_F: dummy variable for gender identification, coded 0 for male and 1 for female

 race: categorical variable, coded 0 for white, 1 for Asian, and 2 for other

 promote: dummy variable indicating if employee was promoted within the past 11-months

There are no missing data in this final version of the data set.

In addition to these concrete variables, the employees also completed the “Personality Questionnaire”.

Items on this self-rated instrument asked how well the following word/phrase describes the employee

on a 1 (not at all like me) to 5 (very much like me) scale:

 do a thorough job  reliable  perseveres

 original  imaginative  shy

 reserved quiet  sticks to a plan

 curious  inventive

The ultimate goal of the study is to determine what sort of people get good job ratings and get

promoted. In particular, the clients are interested in the association between personality type and job

ratings. However, they also want to examine any other possible relationships that may or may not

require attention.

For this analysis, there are a few necessary conversions/transformations for the data. First, be sure

that you have coded the race variable as a categorical (factor) variable, or have created the necessary

dummy variables. Please use “white” as the reference group. For gender, no changes are necessary,

but please use “male” as the reference group. As is common with variables such as salary, it is better

to work with the log(salary) instead of the untransformed value. Create a new variable, and use this

log_salary variable in your analyses. Lastly, convert the questionnaire data into a smaller set of

subscales, as indicated next.

First, it is necessary to summarize the data from the personality questionnaire. Conduct a principal

components analysis with promax (non-orthogonal) rotation to determine the number of factors.

Then conduct an exploratory factor analysis on the 11 personality variables to develop summary

measures of personality. (In your final write-up, be sure to indicate ?How many summary measures

are needed to describe personalities? ?How well do these summary measures capture the information

from the 11 personality variables? ?What is measured by these summary measures? and for all of your

answers, ?how did you reach this conclusion?)

Examine all possible bivariate relationships among the data: examine multicollinearity (between pairs

of independent variables) and possible confounding (between indep. vars. and the dependent

variable). Of particular importance are questions regarding whether gender and/or race are related to

the dependent variables (with and without controlling for the possible confounding variables). You

probably will want to explore models that examine the effect of gender after controlling for other

variables.

Based on your decisions to generate scales for the personality measures, run a regression analysis to

predict job rating from all of the available information. Next, run a logistic regression analysis to

predict promotions from all of the available information (including job rating). Construct the best

models (and justify the choice for those models). Assess model fit.

For all dichotomous variable analyses, se sure to report and interpret relevant odds-ratios.