How to Analyze Your Ava Chart Like a Data Scientist
This is a guest post from Ava data scientist Franziska Dammeier.
One of the most common themes in my interaction with Ava users is concern about changes in their Ava charts: Does a decrease in resting pulse rate mean I’m not pregnant? Does an increase in temperature mean I already ovulated? What does it mean if my temperature went up but my resting pulse rate went down?
Waking up to see that one of your data points went down can be stressful, but most of the time, it doesn’t mean what you think it does. Here’s why: when it comes to tracking your cycle, what matters is patterns, and a single data point does not constitute a pattern. Heck, even two or three nights of data isn’t always enough to constitute a pattern.
So how do you know when you have enough data on your chart to get excited or stressed about? It’s complicated. As a data scientist, many people think that I have the answers to everything and can magically look into your chart and tell you if you’re pregnant. But actually, I think one of the most important advantages a data scientist has is patience and skepticism. In other words, if you ask me if you should be excited or stressed about your chart, most of the time I’ll say neither, and tell you to wait a few more days.
Let’s walk through a few examples.
Early rising pulse rate = early ovulation?
In the chart on the right, we can see that resting pulse rate begins to increase as soon as the user’s period is over. Since an increased resting pulse rate indicates that the body is gearing up for ovulation, does this mean that the fertile window is already open, and ovulation is going to occur earlier than predicted?
Answer: it’s too early to tell! While it is, of course, possible to start your fertile window early in your cycle (and even to get pregnant on your period), if this is earlier than you typically ovulate, it’s best to be patient and not jump to conclusions so early in your cycle.
If you wait a few days, the chart will probably look something like this:
You can see that while it looks like resting pulse rate rises significantly very early in the cycle, once you have more data, that initial rise doesn’t look so significant after all. With more data you can see that, overall, pulse rate stays in a limited window and didn’t increase significantly until later in the cycle.
Decreasing temperature: not pregnant?
Here we see a nice temperature increase after ovulation, but at eight days past ovulation (DPO), there is a dip in temperature. What does that mean for your chances of pregnancy this cycle?
It’s too early for it to mean anything. The first thing to note is that it’s probably too early for implantation to have occurred yet. (Most often, implantation occurs on 9DPO). And since implantation of the embryo is what causes progesterone levels to remain elevated, which in turn causes temperatures to remain elevated in early pregnancy, the fact that temperature dropped a bit in the days before implantation usually occurs probably is not a significant sign of anything.
It’s also important to note that one or two nights of decreasing temperatures aren’t significant on their own, even if they occur after the implantation window. As with anything that involves data science, a single data point on its own is not usually significant. Before you come to any conclusions, you need to wait until you have enough data to see a real pattern.
The graph above shows one possible outcome a few nights after the initial decrease in temperature. In this chart, temperature continues to drop for several days after the initial dip. This chart indicates that the user is unlikely to be pregnant and her period is imminent.
The graph above shows another possible outcome after the user collected a few more nights of data. Here, the initial decrease in temperature was just a small dip, and while temperature fluctuates, it does so at a high level, never decreasing significantly.
Is this user pregnant? It’s still too early to tell. The temperature might still decrease significantly in a few days, followed by her period. Or it might remain elevated until after her period has already started. It’s also possible that temperature will remain elevated and she will get a positive pregnancy test.
If it’s always too early to tell, what can you tell from your chart?
The most important data science lesson I can impart to Ava users is that anything new that happened on your chart today should be taken with a big grain of salt. It’s best to wait three more days—by then, you will have a much better sense of whether the change you saw was an outlier or part of a larger trend.
Of course, sometimes you might see what looks like a trend over three or four days, only to find that after another week or so, more data again changes the whole story. This is one of the reasons why Ava works better after you’ve worn it for a few months—over time, it becomes better at differentiating between real and fake trends while they are happening.
Humans are very good at finding patterns in data—in fact, we are too good at it, often finding patterns where none exist! When we think we’ve found a pattern, suddenly everything we see in the world seems to fit perfectly. This is where algorithms come in handy. While algorithms are not perfect, they have some advantages over humans when it comes to pattern recognition. For one thing, algorithms are not influenced by emotion, allowing them to look for patterns with less bias. And for another, they can handle data in many dimensions, allowing them to discern patterns that humans might miss.