Biostatistics Assignment
This assignment is a computer-based exercise using a teaching dataset downloaded from the "Framingham Heart Study".
Background -
The Framingham heart study is a longitudinal study with the objective to study the aetiology of cardiovascular disease (CVD). This study followed up a large group of participants since 1948 who lived in Framingham, Massachusetts, USA, who were free from CVD at the beginning of the study. The original cohort contains 5,209 residents aged between 30 and 62. The participants have been physically examined every other year to collect their medical data as well as having lifestyle interviews.
During the past 69 years, much research has been conducted based on these data. This has considerably advanced our knowledge on identifying major risk factors of CVD.
For simplicity, we ignore the "longitudinal" element in the original study design. Therefore, all the questions/inferences made in this assignment are confined to this assignment and will not necessarily reflect the "facts" we already know about CVD.
Question 1 -
1.1) How many records are there in the dataset ("frmgham2.csv")? How many participants are there in the teaching dataset? Explain why these two numbers are different.
1.2) Create a file that contains only records related to the First Examination ("PERIOD=1") and call it "DATASET1". How many subjects undergo the first examination? Note that, all the questions below (from (Q1.2) onward) will be based on "DATASET1" only.
1.3) Design a table with relevant summary/descriptive statistics, stratified by gender ("SEX"), to describe the population enrolled in the first examination. Write a paragraph (no more than 250 words) to describe the data and any findings from this table. This table should contain the following variables: "AGE", "BMI","SYSBP", "DIABP", "CURSMOKE", "CIGPDAY", "DIABETES".
Question 2 -
In Q2 we are interested in using the First Examination data ("PERIOD=1" or "DATASET1") to study the relationship between a subject's "BMI" and Hypertension ("PREVHYP").
2.1) Examine the distributions of the variable "BMI" for those who have Hypertension and those who are Hypertension-free using histograms. Include these in your report. Describe the distribution of this variable.
2.2) Generate a box and whisker plot of "BMI" for those who have Hypertension and those who are Hypertension-free at First Examination. Include the boxplots in your answer. Describe different aspects of the box and whisker plot.
2.3) Using the histogram from (2.1) and boxplots from (2.2), state the relationship between Hypertension and BMI?
2.4) Conduct a statistical analysis to study the relationship between having Hypertension and BMI. State your hypotheses and write up your conclusions.
2.5) From your statistical investigation of the relationship between BMI and Hypertension above, can you conclude that Hypertension (or no-hypertension) is induced by high (or low) BMI? Explain your reason.
Question 3 -
In Q3 we are interested in the First Examination data only ("PERIOD=1" or "DATASET1")
Using the values of BMI, one can categorize a subject into "underweight", "normal", "overweight", "obese" (4 groups). Create a new variable "BMIGP" using the definitions below:
- "BMIGP=1": If one's BMI<18.5, it falls within the underweight range.
- "BMIGP=2": If one's 18.5≤BMI<25, it falls within the normal range.
- "BMIGP=3": If one's 25≤BMI<30, it falls within the overweight range.
- "BMIGP=4": If one's BMI ≥ 30, it falls within the obese range.
3.1) Display the frequency table of the new variable "BMIGP". Include any missing values on your table if there are any.
3.2) Cross-tabulate variables "BMIGP" with smoking status ("CURSMOKE"). What is the prevalence of smoking in each BMI group? What can you observe from these results in terms of the relationship between Smoking status and BMI?
3.3) Conduct an analysis to study the relationship between smoking status and BMI groups. State your hypotheses and interpret your findings.
Question 4 -
In Q4 we are using the First Examination data only ("PERIOD=1" or "DATASET1") to study the relationship between Systolic Blood Pressure ("SYSBP") and BMI.
4.1) Use any software to generate a scatterplot of "SYSBP" (on Y-axis) and "BMI" (on X-axis). Include the plot in the report. Calculate Pearson correlation coefficient. Use visual inspection as well as Pearson coefficient to describe the relationship between BMI and SYSBP.
4.2) Assume Simple Linear Regression (SLR) analysis is to be used to study the relationship between Systolic BP and BMI. Treat "BMI" as the independent variable and "SYSBP" as the dependent variable.
4.2.1) Write down the model/formula of SLR in term of "BMI" and "SYSBP".
4.2.2) State the estimation method used to find the best linear fit of the data.
4.2.3) List the assumptions used behind SLR for SLR to be a valid model for data analysis.
4.2.4) Examine all the assumptions listed in (4.2.3) using "SYSBP/BMI" in "DATASET1". Report whether each assumption has been satisfied or not and justify your answer. Hint: Plot the boxplots of "SYSBP" stratified by "BMIGP" and include these in your report. These boxplots together with the scatterplot generated in (4.1) may help you to justify some of the assumptions.
4.2.5) Conduct SLR using any computer software. Write out the estimated regression line. Is the relationship between BMI and Systolic BP statistically significant at the 5% level? Justify your answer. Use the estimated coefficient(s) to explain the size of effect between BMI and Systolic BP.
Attachment:- Assignment Files.rar