New version page

# UW BIOST 558 - Project Proposal

Pages: 4
Unformatted text preview:

​ Heart Disease Mortality Data Among US Adults (35+) by State/Territory and County – 2015-2017Project ProposalProject descriptionWe will analyze heart-disease mortality data for the United States. The response variable will be the mean mortality rate. The analysis will use a factorial design that compares mean mortality rates using factors forstate, county, gender, and race/ethnicity.Data cleaning and preparationWe expect to do some data cleaning and preparation before we begin our analysis.Analysis questionsGlobal (across US) questions● Across the United States, is the mean mortality rate the same between genders?Analysis plan:1. Analyze variance and look for any outliers. 2. We will use a Welch t-test because the normality assumption is not met for our dataset and there is a difference in variance. This is based on a sample size of 32231.a. Variance of men is twice that of women at the county level.3. Verify the power of the test using simulation. Our target power is 0.9 and target significance level is 0.05.● Across the United States, is the mean mortality rate the same between different races/ethnicities?Analysis plan:1. Analyze variance and look for any outliers. 2. We will use a pairwise 5 t-test. We will use these to determine whether there is a difference in means using the Bonferroni correction.3. We also considered using three races/ethnicities that have the most data if we find that the dataset does not contain sufficient data for all of the races/ethnicities specified.4. Verify the power of the test using simulation. Our target power is 0.9 and target significance level is 0.01 for the individual tests because of the Bonferroni correction.Questions on controlling for the effects of different factors● Across the United States, do we see the same differences between genders if we control for race/ethnicity and vice versa?Analysis plan:1. Analyze variance and look for any outliers. 2. Use linear regression with all five races/ethnicities and the two genders (male and female) to see if controlling for each of these factors changes the effect of the others.● Across the United States, does controlling for state change the effect of race/ethnicity and gender on mortality rates? Analysis plan:1. Analyze variance and look for any outliers. PA2. Use linear regression to analyze whether Washington state has the same mortality as the nation as a whole.3. Use linear regression to analyze whether Washington has the same mortality as the nation overall when controlling for 1) the two races for which we have the most data and 2) the two genders (male and female).PAData descriptionWe are using the following dataset from the Centers for Disease Control (CDC). Heart Disease Mortality Data Among US Adults (35+) by State/Territory and County – 2015-2017 The following table describes the subset of columns from the dataset that we will use for our analysis.References:We’ve been looking at other experiments and results that have been published to understand whether they are consistent with our results.https://www.cdc.gov/heartdisease/facts.htmhttps://www.health.harvard.edu/heart-health/the-heart-attack-gender-gapPAAppendix:● Is the mean mortality rate same between different states in the United States? If not, which states showthe greatest differences?Nested (within state) questions● Within each state, is the mean mortality rate the same between genders? If not, which states show the greatest differences.● Within each state, is the mean mortality rate the same between different races/ethnicities? If not, which states show the greatest differences.● Within each state, is the mean mortality rate the same between counties? If not, which states show the greatest

View Full Document