Statistical Analysis Using Stata 15 Assignment Answer

Introduction

Get free samples written by our Top-Notch subject experts for taking online assignment help services 

The data set is from a randomized control trial, consisting of 1467 school children aged between 5-9 years. Data was collected at 3 time points; hence I will be analyzing the data at all 3 time points/cross-sections. The aim is to see if there is any association between waist circumference and blood pressure in children, and if there is any does it vary with ethnicity. There are 2 primary outcomes- systolic blood pressure and diastolic blood pressure (continuous variables); two secondary outcomes systolic and diastolic blood pressure dichotomized as normal or elevated.

Analysis

Creation of new variables

The first step of the analysis entailed the creation of new variables which dichotomized the systolic bp z scores and the diastolic bp z scores into normal and elevated. The 90^th percentile was used to determine whether a score is normal or elevated. Scores which were equal to or less that the 90^th percentile were normal while the rest were taken to be elevated. Dichotomous variables were generated for all the 3 points measured. The variables were then named;

• p_mebl_sys_dicho
• p_mebl_dia_dicho
• p_mef1_dia_dicho
• p_mef1_sys_dicho
• p_mef2_dia_dicho
• p_mef2_sys_dicho

Descriptive statistics

We also performed descriptive statistics to investigate how the variables of interested changed within the 3 points i.e. at baseline, follow up 2 and follow up 3. Results obtained are as shown below;

Linear modeling (Model 1)

Linear modeling was done to investigate the effect of explanatory variables on the dependent variables which are systolic blood pressure and diastolic blood pressure. The explanatory variables for the linear models formed are waist circumference, baby weight, height, sodium intake, minutes of MVPA and sedentary time. Results for the analysis done can be categorized into baseline, follow up 2 and follow up 3.

Baseline

Analysis of baseline data showed that weight, sodium intake and school id are the variables which proved to be statistically significant predictors of systolic blood pressure with p-values less than 0.05. The coefficient values represented in Table 1 and 2 show the coefficient values of the different predictors which can be taken to be the values by which the predictors either increase (if positive) or decrease (if negative) for every unit increase of the dependent variable. With the waist score adjusted for sex, we see that for every unit increase in systolic blood pressure, the waist score in females increase by 0.07. On the other hand, for every unit increase of the systolic pressure, height decreases by 0.01 while weight increases by 0.052. For the diastolic blood pressure, for every unit increase, the waist score for females increased by 0.103. School ID and female waist circumference were identified as the statistically significant predictors for diastolic pressure since they had p-values which are less than 0.05. The data proves that gender had greater effect on diastolic BP when compared to systolic BP.

Follow up 1

For the follow up 2 study, the only variables identified as a statistically significant predictors for both systolic blood pressure and diastolic blood pressure are the baby weight and the school ID which had a p-value of 0.000 and 0.009 respectively. None of the predictors was however statistically significant for the diastolic blood pressure. Table 3 and 4 shows the coefficients for the predictors for the follow up 2. The coefficients represent values by which the predictors either increase (if positive) or decrease (if negative) for every unit increase of the dependent variable.For the first follow up, female waist circumferences were seen to decrease with single unit increase in blood pressure.

Follow up 2

The follow up 3 represents the 3^rd point of the study, baby weight and height were still identified as the only statistically significant predictors for the systolic blood pressure. For diastolic blood pressure, the weight has a p-value of 0.000 showing that it is a statistically significant predictor of the dependent variable which is diastolic blood pressure. Table 5 and 6 shows the coefficients for the predictors for the follow up 3. The coefficients represent values by which the predictors either increase (if positive) or decrease (if negative) for every unit increase of the dependent variable.

Model 2 (Linear modeling with interaction term)

Table 7 to12 show the coefficients for the various predictor variables above with an interaction term of ethnicity introduced to the waist circumference. Waist circumference was a statistically significant predictor of systolic BP amongst Asians. The waist circumference was also statistically significant predictor of systolic BP amongst ethnicities in the group others. For the diastolic BP, waist circumference in Asians, weight and school Id were identified to be statistically significant predictors. For diastolic BP in the first follow up, school Id and weight have been found to be statistically significant predictors having 0.003 and 0.001 p-values respectively.

Model 3 (Logistic regression)

For the baseline study, the sodium intake was identified as the only statistically significant predictor in categorizing both systolic and diastolic blood pressures. For the follow up 2, baby weight was identified to be statistically significant covariate for systolic blood pressure but not diastolic blood pressure. The follow up 3 found no statistically significant covariates for both the diastolic and the systolic blood pressures. The base outcome is the "normal” blood pressure. The logistic regression tables give the coefficients for the different variables which represent the odds ratio for the different variables. Depending on the sign on the coefficient value there is either an increase (if positive) or a decrease (if negative) in the odds of getting an elevated blood pressure for a single unit of the covariate e.g. for our statistically significant covariate in the baseline, we can say interpret the results us; holding other factors constant, there is a 0.0002 increase in the odds of getting an elevated blood pressure for every single unit increase in the intake of sodium.

Model 4 (Logistic model with interaction term)

For the 4^th model we mainly studied the follow up 3 by first modeling without the interaction term then creating a second model with an interaction term to assess the difference brought forth by the interaction term. The covariates used were waist circumference and ethnicity. The results obtained show that waist circumference is statistically significant predictor with a p-value of 0.000 which is less than 0.05. Although the ethnicity variable is not a statistically significant variable, the introduction of the introduction term improves its significant since it reduces its p-value for both the diastolic and systolic blood pressure. Tables for model 4 give coefficient values which represent odds ratios with normal BP as the base outcome. Depending on the sign on the coefficient value there is either an increase (if positive) or a decrease (if negative) in the odds of getting an elevated blood pressure for a single unit of the covariate