B5. In this analysis, you will be running a Kruskal-Wallis test, comparing women in the four BMI categories (bmicat)

with regard to the number of miscarriages they had ever had (miscarr). Number of miscarriages in a ratio-level variable, but it is skewed, with most women having had no miscarriages. Begin by running a simple Crosstabs (following steps described previously) to examine the distribution of miscarriages among women in the four BMI groups. Then run the Kruskal-Wallis test via Analyze ➜

Nonparametric ➜ K Independent Samples Test. Move miscarr into the slot for Test Variable List, and bmicat into the Grouping Variable slot. Define the range for the groups as 1 (Minimum) to 4 (Maximum). Select Kruskal-Wallis as the Test Type. Click the Options pushbutton and request Descriptive Statistics. Then click Continue, then OK, and answer these questions:

(a) What percentage of women had had one or more miscarriage in this sample? (b)

Which BMI group had the highest incidence of a miscarriage? Which group had the lowest incidence?

(c) What was the range of values for the miscarriage variable? (d)

What was the mean number of miscarriages?

(e) Based on the mean ranks within the Kruskal-Wallis test, was the relationship between number of miscarriages and BMI linear—that is, was there an incremental progression of ranks for miscarriage as BMI got larger?

(f) What is the value of the chi-square statistic for this Kruskal-Wallis analysis?

(g) Is this significant for a .05? At what level? (h) What does the result mean? (i) Could we have analyzed these data using a chi-square test? Why or why not? (j) Could we have analyzed these data using ANOVA? Why or why not?