Average Gas Consumption

[dirkhh]

The Subsurface graph of course includes the confidence interval, but "whatever", right?

 

[scubadada]

Hi @Storker

I admit, this is not a great line. Unfortunately, my simple Excel graph does not have CIs. However, one thing missing from my initial graph is any idea of how many observations make up a point in the scattergram. There are 540 data points on the graph. This version attempts to weigh the number of observations at each point. I don't have a legend for the weight of the data points. Nothing fancy, perhaps just a trend.

 

[rx7diver]

The problem with such pictures is, by either stretching or compressing the axes, you can make the line appear to be more, or less, horizontal. Someone above mentioned the coefficient of determination (a.k.a., "R-Square" or "R-Squared" or R^2). This would be an appropriate statistic to consider here.

Looks like there is (at least) one high-leverage point. You might want to recompute things after omitting this point.

rx7diver

 

[rx7diver]

... Regardless, I have fun looking at my dive data when I am not lucky enough to be diving

+1

rx7diver

 

[atdotde]

The problem with such pictures is, by either stretching or compressing the axes, you can make the line appear to be more, or less, horizontal. Someone above mentioned the coefficient of determination (a.k.a., "R-Square" or "R-Squared" or R^2). This would be an appropriate statistic to consider here.
rx7diver

In Subsurface, you can show the 95% confidence interval for the regression line (which is related to R^2).

 

[rx7diver]

In Subsurface, you can show the 95% confidence interval for the regression line (which is related to R^2).

@atdotde,

The coefficient of determination (R^2) will give you the percentage of the variation in the data that is accounted for by the estimated regression line. So, a small R^2 suggests that the regression line should *not* be used either for estimating a *mean* SAC rate (at a given water temp, within the scope of the water temps seen here), or predicting a SAC rate (at a given water temp, within the scope of the water temps seen here), or forecasting a SAC rate (at a given water temp, beyond the scope of the water temps seen here).

If R^2 is appreciable, then the next step is to look at the residuals to ascertain whether the underlying "assumptions" seem satisfied. If the underlying assumptions seem satisfied, then you can do the "formal" stuff--like computing the confidence bands or prediction bands--which depend on the underlying assumptions being satisfied.

What's important here is that, even if software will automatically generate confidence/prediction intervals for you, these CI's/PI's should *not* be believed unless the underlying assumptions are satisfied.

rx7diver

 

[atdotde]

I think, you got small and large R^2 mixed up. A value of 100% means all data points fall exactly on the fit line. If it is small, there is a lot of scatter, the confidence bands will be wide and you could fit many different lines between those. Given how noisy typical diving data is, I would not lose too much sleep over the fine print and rather use those as a qualitative measure of how much of a trend can actually be read out of the data.

 

[rx7diver]

I think, you got small and large R^2 mixed up.

I think you misread what I wrote. (My day job, per my profile description.)

rx7diver

 

[rsingler]

I'm not sure if this helps me have a low SAC rate average but my wife bought a device. Top is supposed to be blood oxygen level and you are supposed to be at 95% or above. Below is heart beats per minute.

She thinks I will soon be a zombie. This taken when I am at work in my office.

View attachment 663364

There are LOTS of reasons why fingertip capillary saturation can be inaccurate.
But if 88% is real on room air, that should be of concern to you. The "shoulder" of the oxyhemoglobin dissociation curve starts around 90%, so any further drop in your carried oxygen may result in a markedly decreased saturation, which is not good for your downstream tissues (e.g., heart and brain).
If you think that number is real, please consider visiting your doctor.

Diving Doc

 

[tursiops]

There are LOTS of reasons why fingertip capillary saturation can be inaccurate.
But if 88% is real on room air, that should be of concern to you. The "shoulder" of the oxyhemoglobin dissociation curve starts around 90%, so any further drop in your carried oxygen may result in a markedly decreased saturation, which is not good for your downstream tissues (e.g., heart and brain).
If you think that number is real, please consider visiting your doctor.

Diving Doc

First thing he ought to do is try a different finger/hand.