Tracking transition with testosterone

Maybe test/track grip strength (each hand) using an inexpensive hand dynamometer?

2 Likes

oh, that sounds much easier. I’ve ordered one, thank you!

Update week 4: emotions and muscles

I’ll probably update less frequently going forward, but this week feels like some significant observations. Also, thanks for ideas from @ejain & @QuantifiedBob I’ve gotten a grip strength tester and did measurements with tape. I already blew away my spouse with grip strength but that’s because he’s got the grip strength of a 14-year-old.

Chris probably assumes I’m going to be stronger than him pretty soon, and he might be right. I’ve had some nights/mornings where I’m feeling muscle pain – like I have the flu, or working out. I’ve felt like it’s a lot easier to lift things. And looking in the mirror, my biceps are really clear. (Then I start posing, and then I remember this animated gif.) It feels like cheating, I haven’t worked out at all.

It’s still a lot harder to cry. (Maybe easier at the end of the week when T is low.) My behavior isn’t so dramatic now when upset, but when I record it, I’m clearly unhappy – like I’m “bottling it in”? Recently I almost never feel “good”. I feel slightly anxious all the time. It might just be some recent events got me down, but look at this mood tracking data – it’s pretty striking! I’m wondering if I need to learn how (some) men handle emotions (everyone is different, this is just “me on androgens” not a universal truth) – sorry, this all sounds like I’m making some misandrist joke, but … yeah. I’ve read some (trans)guys recommend exercise.

I’ve been recording when I inject (day and approximate time), because I wondered if there’s any mood effects.

Today I put together a notebook in Open Humans analyzing my iMoodJournal mood tracking against the injection times.

TLDR: the signal doesn’t seem to be very strong. Maybe there’s something there, but not dramatic and probably not statistically significant. (I’m not clear on how to test for significance, I’m not very expert on analysis.)

Here’s a link to the notebook in case anyone would like to re-use or adapt code from it: https://exploratory.openhumans.org/notebook/75/

And here’s the graph:

If I squint, maybe there’s a shift of one level up, post injection? The mood data is very noisy. There’s also a potential confounder regarding any other weekly cycles (since injections are roughly weekly, although I’ve shifted the day of the week a bit over time).

1 Like

Plotting the averages (and perhaps also the variances) for each day could help make the data a bit less noisy…

Right – I did try one form of averaging, the dotted line is a rolling average with a window size of 8. (Assuming I used that pandas function correctly! I’m pretty novice at using it.)

But… I could increase the window size, that’s easy. Smoothing with n=40, it does look like there’s a mood shift. :thinking:

(I’ve shared the updated notebook with an addendum – thanks for the nudge!)

I’m skeptical because there could be a different weekly rhythm explaining it. So… I’m going to deliberately nudge my injection timing in upcoming weeks! I had been timing injection to Sunday to deliberately aim the “low T” to non-workdays, but – hah – as a result I can’t disentangle causality (mood could be lower just because weekends are exhausting, we have three young kids :sweat_smile:).

I guess there really might be a mood change!

I updated the notebook again, adding the same plot against the start of each Monday to see if there’s a weekend low (I usually inject at the start of the week). That doesn’t seem to be present. So… I don’t think there’s a weekly rhythm that explains it.

Have you tried simply aggregating by day, rather than using a rolling average?

If you were already tracking mood prior to starting the injections, you could check the averages for each weekday (and compare before and after)…

These graphs are available on open humans?

@rain8dome9 I’ve shared the notebook, which can be loaded and run via Open Humans. :slight_smile: a link to that is here: https://exploratory.openhumans.org/notebook/75/

The graphs come with the notebook, although they will be overwritten upon running it again. My mood log data is private, so you can’t run it again on my data. But it could be run on another dataset from iMoodJournal, which can be uploaded here: https://upload.openhumans.org/upload?datatype_id=20

The injection schedule is embedded in the notebook itself and you could replace it with whatever events are of interest. I’m becoming interested in medication/treatment tracking now & wondering what the best tool is to be tracking that.

Could I download the notebook and and run it on my computer without ever uploading data? I am much better with R than python. Can OpenHumans run R notebooks?
Why is injection scheduled embedded rather than data?

Ah, it looks like we didn’t add a link for direct download, but it’s available – https://exploratory.openhumans.org/export-notebook/75/

Yup, Open Humans runs R notebooks too. (Basically, those two languages. Not sure if we support Julia, the third language founding “jupyter”, it’s not nearly as popular.) A recent example @gedankenstuecke made: https://exploratory.openhumans.org/notebook/76/

I could just upload a file, it didn’t seem to matter much either way with such a small list. I think I figured if hard-coded it might be clearer for someone that wants to re-use this and replace with their own target events.

Would there be any interest in adding a regular ‘download’ button for the notebooks to the front end? Most of the notebooks make pretty specific use of the Open Humans API at least for getting the data, but if people want to adapt them for local use we could offer that!

Yep, We have Julia, Python & R kernels available for the notebooks. If there’s interest in additional kernels, that would be possible too, but so far there hasn’t been any request for those.

1 Like

What I really want is to work offline; have a copy of open humans running on my pc without internet at all. And I want this to be easily available to the end user. I mean installation should be fairly easy. It does not have to work on any data but my own.

@rain8dome9 it’s using jupyter & other python packages – you can probably install and run those things locally on your own machine. But, well, writing and updating anything to “fairly easily install” and run locally in various platforms is much more challenging to create, maintain, and update.

Open Humans isn’t a commercial product. It’s a nonprofit supporting a community & platform, for folks that want to share & build projects that work with personal data, as individuals and collective projects. It’s open source, the community chat is open, we’ve started monthly open community calls. Community is also part of governance (e.g. a third of our board seats are elected by members).

If someone wants to donate coding skills (or lots of funding, I guess!) to see more local / offline things integrated somehow, I’m sure we’d be happy to explore that. :slightly_smiling_face:

Took me two days to install a medium complexity (server) python program but that was under unusual circumstances. How about a guide on setting up my own instance? Data never leaving the personal computer is the best security.

There’s plenty of guidance already for installing jupyter. :slight_smile: https://jupyter.org/install

The VMs provided contain a set of packages that notebooks can assume are available. A custom installation might need to add some, depending on the notebook.

Open Humans is definitely whatever people are comfortable with, people have different preferences. In the future I think we’d like to have more first class support for notebooks as shared objects in themselves, rather than add-ons that presume data present within accounts.

Ok that is jupyter notebooks. What about the Open Humans API?

@rain8dome9 I think this got a bit side-tracked here. I’ve shared an updated notebook that has code that lets you run it on local files. (But if you have your own project you’re trying to do re-using something I did, maybe start a new post about that… :slight_smile: )

Usually for LOESS that is called a confidence interval. I think this suggestion over short window could be very productive.