and my colleague and I, Dr. Mary Jean Amon.
For this project, we examine the relationship
between legislative restrictions on the use of public resources for abortions
and their impact on fertility rates based on maternal age in the United States.
We believe that by examining these regulations,
we could offer insight into potential impacts of future legislative changes,
thus aiding and understanding the dynamics
and potential consequences of such policy shifts on fertility rates.
I'll start by quickly discussing our data sources,
and then I'll demonstrate how we use JMP to perform a two-way ANOVA
and a late-in class analysis to investigate these relationships.
There were two data sources used for the study.
The first was legislation that was
obtained from a law, ATLAS policy surveillance program dataset.
The secondly, we got the 2021 fertility rates from CDC's WNDYR database.
The data sets were combined together
based on the mother's state of residence in the year,
with a one year lag in the birth status
so that the legislation was in effect when the pregnancy began.
Our first objective was to identify impacts of restrictions
on public resources on fertility rates.
We broke the legislation down into three buckets,
whether the state had no restrictions,
whether they had restrictions but excluded Medicaid,
and then whether they were restrictions, including Medicaid.
A two-way analysis of variance was performed
to determine whether there was a statistically significant difference
between the mean fertility rates based on these three buckets in the maternal age.
We first started by visualizing these
three categories with the map that we made with JMP's Graph Builder.
From there, we observed that there was
an uneven number of states in each of the restriction categories.
We used a common practice of running the analysis with the rate transform data
to avoid any challenges from the imbalance data.
I'll demonstrate first how we add this ranking to the data set,
and then we'll go through the two-way ANOVA.
First, you start with Analyze, Distribution.
We're going to put in the variable that we want to do the rank transformation on,
which is fertility rates and to Y and hit OK.
We go down to our lovely red triangle then to Save and Ranks.
That's going to save the ranking
of from lowest to highest of the fertility rates onto our main data set.
We're just checking that it's there.
Next, with the results of the rank fertility rates,
we're going to use that to do our two-way ANOVA.
Back to Analyze Fit Model,
we're going to add in not the original fertility rates,
but that ranked fertility rate that we just created to our Y.
We're going to add in our two independent
variables in full factorial, so that we can do that two-way
for the maternal age and that restriction category will hit Run.
Then it gives us the same outputs that we would obviously expect,
but it's using those rank transformations.
We can go down and look at the effect summary to see that there are
statistically significant interactions going on there.
The two-way ANOVA identify that there were statistically significant differences
in those fertility rates based on maternal age and the presence of restrictions.
Specifically, we observed that women ages 15 to 29 had lower fertility rates
when there were no restrictions present in their state.
The next objective was to identify
patterns in the state abortion restriction composition.
We use a late-in-class analysis because we had these binary,
these yes or no indicators for six different categories of interest.
They were related to government funds, government facilities,
and other various programs like state insurance programs for state employees.
We'll demo how we use JMP's clustering to group those states together
based on these six different categories,
and then how we use Graph Builder to help us display the results
and interpret them a little easier.
To run the late-in class analysis,
we're going to analyze clustering, late-in class analysis.
We're going to put in our six binary indicators into our Y.
Then we could adjust the number of clusters and it would run more than
just three, but we're just going to use three for our purposes to keep it simple.
After running it, we have this high-level
characterization about each one of the clusters.
But let's create an additional visualization that'll start helping us
interpret how the states fell out in each one of the three clusters.
To do that, we're going to go to Graph Builder.
We're going to take our state of residence
and put it down on map shape that lets JMP know that we're wanting to make a map.
Then we're going to put our most likely cluster, which is added to the data set
after we run that LCA, and we're going to use that to actually color it.
This gives us a really quick look
at how each of our states that actually had restrictions,
which of the three clusters they fell out into.
Then you're able to use Graph Builder again
to create other visualizations like we did that allow you to compare each
of the states' composition of their restrictions and their fertility rates
using the three clusters that we created with the LCA.
When we did that,
what we observed is that when there were restrictions on multiple types of public
resources, it was often associated with higher fertility rates for those states.
I just want to thank you for viewing the poster session.
The goal was really to demonstrate how we
used JMP to examine the relationships between abortion restrictions that were
targeting public resources and fertility rates across multiple maternal age groups.
By performing the two-way ANOVA on that rank transform data,
we observed that women ages 15 to 29
had lower fertility rates when there were no restrictions.
Through the use of late-in-class analysis
and visually analyzing the results, we observed that restrictions on multiple
categories of public restrictions were associated with higher fertility rates.
Thank you again.
