Using SPSS to Answer questions.

Using our NHANES dataset, let’s look at how weighting our dataset affects hypothesis test. NHANES Dataset (Please Download)

1. The current estimated average family size in the U.S. is 3.23 people. It would be good to know if the NHANES sample is representative of this average. We will be using our One Sample T Test in SPSS (Analyze Compare Means One Sample T Test).
   1. Using an unweighted dataset, compare the average family size (DMDFMSIZ) with the test value of 3.23
      1. Interpret the mean differences, critical value and confidence intervals.
   2. Using a weighted dataset, compare the average family size (DMDFMSIZ) with the test value of 3.23
      1. Interpret the mean differences, critical value and confidence intervals.
   3. Compare the unweighted and weighted sample means.
      1. Why are they different? Which do you think we should be using? Why?
2. The current estimated percentage of people who are overweight in the U.S. is difficult to ascertain (based on the Body Mass Index, or BMI). One way to reduce the bias when asking people their height and weight (people tend to underestimate their weight and overestimate their height) is to measure it directly. In your NHANES dataset, we do not have physical measurement but we do have one question that asks respondents “Has a doctor or other health professional ever told you that you were overweight?” (MCQ080). Based on this variable, let’s compare our sample data with BMI estimates from survey data. Using published sample data, subtracting individuals who are “obese” according to BMI from those who are either “overweight” or “obese” gives us a proportion of 32.2 %. It would be good to know if our NHANES “overweight” question comes close to this estimate. We will be using our One Sample T Test in SPSS (Analyze Compare Means One Sample T Test).
   1. Create a dummy variable for MCQ080 (similar to previous weeks’ process) where “0” = No and “1” = Yes. All other values should be coded as system missing.
   2. Using a weighted dataset, compare the proportion who have been told they are overweight with the test value of 32.2%. Remember, you must enter the test value in SPSS in its decimal form.
      1. Interpret the proportion differences, critical value and confidence intervals.
      2. Is this a good approach for estimating overweight population proportions? Why or why not?