How much does fc max bpm have to change before it stops being noise? The threshold from 210 nights

With 28 nights of baseline, fc max bpm has to change by 7.62% before it stops being noise. With 56 nights, the threshold drops to 5.39%. With 84 nights, it is 4.4%. With 168 nights, it is 3.11%. These numbers come from 210 nights of one body, where this variable has a CV of 18.05%.

What is fc max bpm, and why does it move so much?

fc max bpm is the highest heart rate the ring records in a night. It is given in beats per minute. The median of this series is 115.5 bpm.

It is a jumpy number. The CV of 18.05% tells you how far one night sits from the next, relative to its own level. A high CV means a single night says little. One high peak is not news. It is normal for this variable.

This matters because the eye looks for stories. You see one night with a high value and you assume something happened. With a CV of 18.05%, the likelier answer is the usual variation.

What is a noise threshold?

A noise threshold is the smallest change that separates a signal from the normal variation of that series. If your average moves less than the threshold, you cannot tell it from chance. If it moves more, it starts to deserve a look.

A look, and nothing more. Crossing the threshold is not a verdict. It only says the change no longer fits inside the measured noise.

The threshold depends on how many nights you average. With few nights, the average is still unstable. With more nights, it settles, and a smaller change shows up. That is why the archive keeps several thresholds for the same variable.

What is the fc max bpm threshold for each window?

This is the row from the archive, as it stands. The table comes from ruido.medido.json#fc_max_bpm.

Field	Value
Nights measured	210
Median	115.5 bpm
CV	18.05%
rcv	50.04%
Threshold at 28 nights	7.62%
Threshold at 56 nights	5.39%
Threshold at 84 nights	4.4%
Threshold at 168 nights	3.11%

The reading is simple. As the window grows, the threshold falls: from 7.62% at 28 nights to 3.11% at 168 nights. An average over more nights is less swayed by one odd night. So it can pick up smaller changes.

The rcv of 50.04% is another measure of spread in the same series. We publish it next to the rest so the row is complete.

How was it measured?

There are 210 nights from one person, on one device. Each night adds one reading. The threshold is computed on the person's own series. It is not compared with anyone else.

A hard rule sits behind this. We never compare the absolute value of two people measured on different devices. Sensor, window and algorithm change from one device to another, and you cannot pull the body apart from the instrument. What we can compare is each person's variability against themselves.

So this threshold is one body against its own history. It is not a norm for a population. It is a method anyone can repeat with their own nights.

This variable is one of the 31 that the archive measures. Each has its own CV and its own thresholds. None inherits the numbers of another.

What gap does this leave in your week?

Without a baseline, a change is not a reading. It is an impression.

That gap is the same one you have with any habit you start this week. A new schedule, a new routine, a new product. If you did not write down how your own series moved before, you have nothing to compare against afterward. You see a different number and you attach a story to it.

You do not need to believe anything from the archive to see the gap. Look at your own series of nights. How much did it move before you started? If you do not know, any later change stays unmeasured.

The instrument that closes that gap is modest. It is a long series of your own nights and a threshold per window. With 28 nights you already have a first reference. With more nights, the reference sharpens.

The limit

This section states what this data does NOT say.

• It is one body. It is 210 nights from one person. Your threshold may differ.
• It does not say what caused a change. A change that crosses the threshold points to no cause. It only says the change is no longer noise.
• It does not say whether a change is good or bad. The threshold does not judge. It only separates.
• It is one device's reading. Another sensor would measure something else under the same name. The proxy is not the thing.
• It is not medical advice. Crossing the threshold is not a diagnosis or a recommendation.
• It applies only with a full window. If you have fewer than 28 nights, the 28-night threshold does not serve you yet.
• It is not a verdict on any habit or product. What is measured here is the noise, not an effect.

One more limit. A low threshold does not make a variable useful. It only says how far it must move to show. Whether it is worth following is a different question, and this piece does not answer it.

The question this piece does not answer

Knowing the threshold is half the work. The other half is looking at a real change and asking whether something specific caused it. That question needs another design: a baseline, an intervention and a rule for saying it did not work.

This piece also does not say which of the 31 variables have a threshold so high that no realistic change would cross it. That is the gap the next reading of the archive inherits.

Keep reading

• Mental Work — all protocols
• related file: magnesium-for-sleep
• open science — the dataset

Sources

• Own dataset: ruido.medido.json#fc_max_bpm, 210 nights of one body. Provenance with a cryptographic hash in corpus.huella.json.