In Part 1 of this series on mood charts, I listed a few tools for tracking moods and explained how mood charts can help you visually evaluate the status of your mood swings. This time, we’re going to have a quick and relatively painless lesson on useful statistics to get an objective analytical view on what’s going on inside your head. The formulas included in orange below are the ones you would use in Excel (ranges are sets of values like A1:A23).
You can run some simple statistics on mood chart data to look for relevant patterns. These are a nice objective indicator that helps me face reality when I don’t want to: the numbers don’t lie, especially when I use Moodscope, because I can’t easily manipulate those scores. Most people who aren’t complete and utter nerds like me don’t know how to do this or interpret the results, so I’m going to give you a quick guide to three stats you can run on mood chart scores to better understand what’s going on in your head. Isn’t that awesome?!? Seriously though, knowledge is power when it comes to managing mental health, and this is one way to gain more self-knowledge.
Pro-tip: The easiest and most valuable thing you can do for improving the meaningfulness of most statistics is to maintain precision. Keep consistent mood data and accumulate a lot of it, use a more discriminating scale like Moodscope, and track your sleep to the half-hour or quarter-hour. If you’re just starting out with mood charting, track everything you can think of: drugs (frequency and type), alcohol (to the serving, based on medical guidelines of serving size), medications (to the mg), sleep (always), periods (for the ladies), exercise (step counts, types, frequency, duration), sex (endorphins = mood support!), and subjective, qualitative details that are better kept in a mood journal. The bigger the spreadsheet, the better, even if it seems like a nuisance. Eventually you can use the stats described below to weed the variables down to the ones that really matter, and then adjust your behaviors accordingly. If you already know advanced stats, you can use the whole shebang for a stepwise multiple regression and get a pretty sweet model of your brain on X, Y, and Z.
So with all of the stats, aside from periods or medications (unless you have adherence issues), the obvious thing to do first is look for averages [=AVERAGE(range)] to get a sense of where things usually are. I like to do this in time-chunks according to meaningful date ranges, or even just month by month to see how the shift of seasons is treating me. For example, I would calculate an average mood score for the time before I went to France, while I was in France, and after coming home from France, to see how being in France affected my mood (I was in France for a month last summer, see point #3 in the prior post.)
The next thing I look at is standard deviations [=STDEV.P(range)]. These measure how much variability there is in a set of values, and if you’re bipolar, I think they’re crucial to examine, since the degree of variability you experience is what it’s really all about. The number you get from the standard deviation is what you add to and subtract from the average to find the usual range of values. I calculated monthly standard deviations on my mood scores to go with averages because it helps me put the averages into context. Here’s an example of my Moodscope averages and standard deviations over the last 6 months:
- September 2011: avg = 47.97, sd = 9.85
- October 2011: avg = 61.81, sd = 10.43
- November 2011: avg = 53.9, sd = 7.04
- December 2011: avg = 52.74, sd = 9.91
- January 2012: avg = 50.87, sd = 7.92
- February 2012: avg = 52.48, sd = 12.78
To better understand how these statistics are related and what they mean mood-wise, here’s the range of values that the average and standard deviation yield, and how I interpret them:
- September 2011: range = 38.12 – 57.82, interpretation = Fairly depressed, rarely above middle and often pretty well below. Sad panda.
- October 2011: range = 51.38 – 72.24, interpretation = Hypomanic much? Oh yes. Yes, yes, yes!
- November 2011: range = 46.86 – 60.94, interpretation = Whoa, almost normal! Meds must be kicking in.
- December 2011: range = 42.83 – 62.65, interpretation = Not as good as November, but not half bad.
- January 2012: range = 42.95 – 58.79, interpretation = Better than December, but not quite as good as November.
- February 2012: range = 39.7 – 65.26, interpretation = WTF happened to destabilize everything? (answer: travel + bad juju med change)
In short, what I really want for mood stability is a score range of around 45 – 60, though you should use your own baseline score +/- 5-10 in each direction. For me, going below 40 isn’t a good sign, and above 65 isn’t necessarily bad but I consider it a warning to pay attention. I find that above 70, there’s serious hypo/mania action going on. Can you see how the average values and variability are reflected in the graphs? (if not, don’t worry, it takes some practice!) I find that it’s easier to interpret the numbers effectively when I have a graph to look at, and vice versa. Throw in some mood journal entries, and you have a pretty good data set from which to draw solid conclusions about what’s been tweaking your moods.
So those were beginner descriptive statistics. What about relationships? All you need is simple correlation [=CORREL(range1,range2)] to see what variables are related. Correlation basically means that when one variable goes up, the other does too, or else when one goes up, the other goes down. It’s not the same as causality, so all we can determine with a correlation is that these things are related, not which one has the effect on the other. That said, sometimes it’s obvious. For me, a short night’s sleep nearly always leads to an increased mood score the next day.
There are other statistical tests that verify significance, which would show whether the correlations are due to chance rather than a meaningful relationship. Those stats get into a lot more detail, however, and I think it’s probably best to draw the line with correlations for now. Let me assure you, however, that the correlations I’m about to discuss are incredibly, overwhelmingly, undeniably, absurdly strong relationships.
So let’s look at a few correlation coefficients to see what’s going on with those moods and sleep. To really understand these nuances, you would need to see the averages and standard deviations for the amount of sleep I get as well as my moods, but for the sake of example, I’m going to keep it simple. Correlation coefficients are relationship measures, so the numbers here refer to Moodscope scores and the number of hours of sleep I got the night before. This gives me a sense of how much sleep and mood are related to one another.
- September 2011: corr = 0.13
- October 2011: corr = -0.50
- November 2011: corr = 0.25
- December 2011: corr = 0.09
- January 2012: corr = 0.06
- February 2012: corr = -0.07
OK, so these are a little less obvious to interpret. Correlation coefficients always fall between -1 and 1. When values are closer to 0, it means there’s less of a relationship. When they’re closer to -1, there’s a negative relationship (more sleep = worse mood) and when they’re closer to 1, there’s a positive relationship (more sleep = better mood). For my purposes, the closer the values are to 0, the more “normal” things are. October was a hypomanic month, as noted above, and confirmed here: the strong negative relationship, along with the other numbers above, means that I was not sleeping much and feeling great! Things started to stabilize in November, when I started Lamictal. Getting more sleep was making me feel better, and since then, it’s been pretty stable, but then again, my sleep patterns and moods have also been a lot more stable overall.
Moodscope gives a 100-point scale, which means doing statistics with those figures will be much more precise than the 9-point scale of my paper mood chart. For that one, I get fairly similar correlation coefficients, but they are rarely verifiable as non-random. One time that the paper mood score was actually significant and the correlation values for both types of mood scores matched up quite closely was October of 2011, so it does occasionally happen when the data are super-closely related.
So that’s it for some basic statistics that you can do with mood charts to better understand what’s going on in your head. I find this kind of analysis makes me face the reality that I really am bipolar. I just can’t deny it when I look at the data this way. It also helps verify my sometimes flawed sense of how stable or labile things are, and was the main tool that helped me identify the things that most regularly affect my mood swings (sleep and meds.) It might seem overwhelming at first, but keeping a mood chart and even playing with the data can give you some real insight into your mental health.