SPSS Analysis: Relationship Between HgbA1c And Weight In Diabetic Boys - Essay.

Regression analysis of HgbA1c and weight at baseline

1. A group of 10-year-old boys were first ascertained in a camp for diabetic boys.  They had their first weight measured at baseline and again when they returned to camp one year later.  Each time, a serum sample was obtained from which a determination of haemoglobin A1c (HgbA1c) was made.  HgbA1c (also called glycosylated haemoglobin) is routinely used to monitor compliance with taking insulin injections.  Usually, the poorer the compliance, the higher the HgbA1c level.  The hypothesis is that the level of HgbA1c is related to the weight.  The data are entered into an SPSS file called “assignment2_diabetes2018.sav”.  Download this file and use SPSS to answer the following questions:

a) Is HgbA1c a useful predictor for weight at baseline? Run a regression analysis with wgt1 as the dependent and HgbA1c_1 as the independent.  Does HgbA1c significantly predict weight?  Give the statistical evidence for your conclusion.  How well does HgbA1c explain the variability in weight in these data?  Give the statistical evidence for your conclusion.  Write a one-sentence summary of this analysis.

b) Repeat the above analysis for HgbA1c one year later.  Giving the statistical evidence, do your conclusions change?

c) Using transform -> compute, calculate 2 new variables:  wgtdiff = wgt2 – wgt1, which measures the change in weight from baseline to one year follow-up, and HgbA1cDiff = HgbA1c_2 – HgbA1c_1, which measures the change in HgbA1c from baseline to one year follow-up.  Use regression analysis to investigate whether or not the change in HgbA1c predicts the change in weight.  What is the statistical significance of the relationship?  How well does change in HgbA1c explain change in weight?  What is the direction of the relationship?

d) Using the results from all three of the analyses above, write a paragraph giving your conclusions, suitable for the conclusions/ discussions section of a paper. What further investigations would you suggest, based on these data?

2. Use the Framingham data set, with period = 1, to answer the following questions.

a. Does mean age (AGE) differ between those on antihypertensives (BPMEDS = 1) and those not on antihypertensives (BPMEDS = 0)?

b. Does mean systolic BP (SYSBP) differ between those on antihypertensives (BPMEDS = 1) and those not on antihypertensives (BPMEDS = 0)?

c. Does mean diastolic BP (DIABP) differ between those on antihypertensives (BPMEDS = 1) and those not on antihypertensives (BPMEDS = 0)?

For each analysis, summarize your findings in 1-2 sentences.

3. A nutrition expert is examining a weight-loss program to evaluate its effectiveness (i.e. whether participants lose weight on the program). Ten subjects are randomly selected for the investigation.  Each subject’s initial weight is recorded, they follow the program for 6 weeks, and they are weighed again.  The data are;

Use SPSS to test if there is evidence of a significant weight loss.  Run the appropriate test at the 5% level of significance.  Write a 1-sentence summary of your analysis.

4. The following data reflect the ages of students at completion of year 7. Test if there is a significant difference in the mean age at completion of year 7 for rural, suburban and urban students using SPSS. Write a 1-2 sentence summary of your findings.

Regression analysis of HgbA1c and weight at baseline

a. Predictors: (Constant), HgbA1c_1

The value of R-Squared is 0.042; this implies that only 4.2% of the variation in weight is explained by HgbA1c.

a. Dependent Variable: wgt1

b. Predictors: (Constant), HgbA1c_1

The coefficients table below shows that the p-value for the HgbA1c is 0.465; this value is higher than the 5% level of significance. The null hypothesis is not rejected suggesting that HgbA1c does not significantly predict weight.

a. Dependent Variable: wgt1

In summary, only a small proportion of variation (4.2%) in weight is explained by HgbA1c and also results showed that the independent variable (HgbA1c) does not significantly predict weight.

a. Predictors: (Constant), HgbA1c_2

a. Dependent Variable: wgt2

b. Predictors: (Constant), HgbA1c_2

a. Dependent Variable: wgt2

One year later the results do not significantly change. Only 4.7% of the variation in weight is explained by HgbA1c. Again, the p-value for the HgbA1c is found to be 0.436; this value is higher than the 5% level of significance leading acceptance of the null hypothesis hence implying that HgbA1c does not significantly predict weight.

a. Predictors: (Constant), HgbA1cDiff

The value of R-Squared is 0.213; this means that 21.3% of the variation in difference in weight is explained by difference in HgbA1c.

a. Dependent Variable: wgtdiff

b. Predictors: (Constant), HgbA1cDiff

a. Dependent Variable: wgtdiff

The p-value for the difference in HgbA1c is 0.083 (a value less than 10% level of significance), hence we can say that difference in HgbA1c significantly predicts difference weight at 10% level of significance.

d. The results obtained showed that HgbA1c does not significantly predict weight. This is true even after one year. However, when we obtain the difference in weight and the difference in HgbA1c for the two periods we saw some improvements. The difference in HgbA1c was able to significantly predict the difference in weight though at 10% level of significance. Further investigation should focus on controlling for time and seeing how the results behave.

We performed an independent t-test was in order to compare the average age for those on antihypertensive (BPMEDS = 1) and those not on antihypertensive (BPMEDS = 0). Results showed that the average age for those on antihypertensive (M = 57.63, SD = 5.99, N = 41) was significantly different with the average age (M = 53.90, SD = 8.07, N = 603), t (642) = -2.900, p < .05, two-tailed. Those on antihypertensive were significantly older than those not on antihypertensive.

We performed an independent t-test was in order to compare the mean systolic BP for those on antihypertensive (BPMEDS = 1) and those not on antihypertensive (BPMEDS = 0). Results showed that the mean systolic BP for those on antihypertensive (M = 171.95, SD = 30.08, N = 41) was significantly different with the mean systolic BP (M = 141.69, SD = 25.34, N = 603), t (642) = -7.305, p < .05, two-tailed. Those on antihypertensive had higher mean systolic BP than those not on antihypertensive.

We performed an independent t-test was in order to compare the mean diastolic BP for those on antihypertensive (BPMEDS = 1) and those not on antihypertensive (BPMEDS = 0). Results showed that the mean diastolic BP for those on antihypertensive (M = 97.39, SD = 14.44, N = 41) was significantly different with the mean diastolic BP (M = 86.27, SD = 13.73, N = 603), t (642) = -5.001, p < .05, two-tailed. Those on antihypertensive had higher mean diastolic BP than those not on antihypertensive.

A paired-samples t-test was conducted to compare initial weight and final weight of individuals after a weight-loss program. There was a significant difference in the initial weight (M = 151.70, SD = 27.43) and final weight (M = 145.10, SD = 24.86) conditions; t(9) = 3.589, p = 0.006. These results suggest that the weight-loss program really does have an effect weight of individuals. Specifically, our results suggest that before the weight-loss program, the individual’s weight more as compared after the weight loss program. 

The mean difference is significant at the 0.05 level.

First, we checked for the homogeneity of variances where we observed the variances to be homogenous (p = 0.148). For the ANOVA test, results showed that there is significant difference in the mean age at completion of year 7 for at least one region. Post-hoc test using Bonferroni showed that significance difference exists in the mean age at completion of year 7 for the urban and Sub-Urban regions.