The T-Test

On Wednesday we used a T-Test to see if there was a difference in mean labor time between women who worked as a doula for income and those that worked as a doula for hobby. The next obvious question is, what is a T-Test and why did we choose to use it?

Remember back to last week when we talked about a chi-square?  With a chi-square we were able to see if two groups differed on a characteristic that was a categorical variable. This means we could look for differences that are basically yes or no categories.  For example, those who had gestational diabetes and those who did not or those with an intact perinium and those without. This works well for some things, but what if you wanted to know how big the difference was?

For example, lets say you broke your population into two groups and with a chi-square you found these groups differed in the number of women who stayed within the IOM recommended weight gain ranges. But this doesn’t help you know how much weight the women gained.  To do this, you need a test that lets you look at the actual amount of weight gained, and the T-Test lets you do that.

In our example, you could use the T-Test to compare the average weight gain for each group to see how different the weight gain actually was. This can be especially helpful if you have two groups that both gain more than recommended.  It can also be helpful if you want to find out if women in centering pregnancy gain a different amount of weight than women in standard care.  Just like a chi-square, the T-Test will give us a p-value.  But unlike the chi-square, it will also give us a confidence interval.

A Literature Example

T-Tests are often used to assess the similarity of two groups before a study begins. It helps to show it is the intervention making a difference, and not an underlying difference in participant ages, weights, or hemoglobin levels.

Check out this study: Social determinants of partner support in pregnancy. These researchers used a T-Test to look at differences on a social support score between two groups. This works better for a T-Test than a chi-square because the social support score has a value that increases with the numbers.  If the social support questionnaire had scored into categories of high, medium, and low rather than a number the chi-square test would have been the better choice.

The Birth Worker Survey

The Birth Worker Survey included questions about the length of labor and the number of years working at particular professions.  We can look for differences in these variables by using a T-Test.

First, we can look at the difference in length of labor in hours by whether or not the respondent works as a childbirth educator.  Using this method we first see the mean for each group.  Those who work as a childbirth educator have a mean length of labor of 5.5 hour.  Those who do not reported a mean length of labor of 8.4 hours.  The mean difference between the two groups was -2.85 with a 95% CI of -9.1 to 3.3. The p-value is 0.352.  This means this data does not give us evidence of a difference in mean length of labor between those who work as childbirth educators and those who do not.

The survey asked childbirth educators how many sessions they have in their typical class.  We can break down the childbirth educators into those for whom it is a job, and those for whom it is a hobby to see if there are differences.  Those who see their childbirth education role as a job have a mean of 7.25 sessions, while those for whom it is a hobby have a mean of 5 sessions with a mean difference of -2.3 sessions.  The T-Test for differences reveals a p-value of .419 and a 95% CI   -7.0 to 2.5.


Jennifer Vanderlaan (Author)