I have heart rate data of an athlete when he was running 35 minutes on a treadmill.
The data were observed every 5s. For time series analysis purposes, how can I convert heart beat time series data? Is this correct for time series analysis?
It is important to note that I am an economist and not an expert in health-related topics such as the heart. However, since no one has answered yet, perhaps I can spark some creativity.
My first inclination is to suggest that no conversion is necessary; that is, all of the heart-beat data could be a series (i.e., variable) related to time. The time interval could be as often as every five second and 15, 20, 30 second intervals also make sense along with a minute or more, The challenge is that this is not all at that interesting without other variables.
If we add additional variables like oxygen concentration in the blood O2 saturation, respiration, blood pressure, temperature, and other things we could measure, it would be more interesting. For example, it would be interesting to identify the lagging and leading variables if the angle of the treadmill was raised periodically throughout the exercise. Let's say the angle increased every five minutes for one minute and then goes back to the original angle. It would be interesting to observe the changes. An interesting question to me would be does the body adapt and anticipate the change (e.g., would the heart rate, respiration, or anything else begin to increase in anticipation of the change in angle?).
Of additional interest might be things like the impact of time. For example, does the individual have a constant heart rate as time increases? Time might be measured like 0 for the time when exercise begins and then perhaps 1 at 5 seconds, 2 at 10 seconds, 12 at 60 seconds, and so forth. I would suggest using a time interval consistent with the variables; my guess, is taking the blood pressure every 5 seconds would be challenging, so if blood pressure is included, then intervals that match it would be preferred. Many of the others could be measured more quickly.
Economists are typically not afraid to make mistakes, so this is my hypothesis:
1. Heart Rate increases rapidly with exercise and then nearly stabilizes with slight increases as more time elapses. Conjecture: Fatigue, dehydration, and similar factors force the heart to work more.
2. Blood pressure (BP) increases rapidly and then stabilizes before a relatively small decrease over time. Conjecture: I would guess the veins and arteries expand slightly with continuous exercise.
3.Respiration similar with BP and I would guess the lungs expand with exercise.
4. Body Temperature. Increases (with the longest lag of any variable and is probably linked with hydration (or dehydration) Eventually, it would stabilize and remain relatively stable.
5 The O2 saturation would decrease and remain lower after prolonged exercise with slightly decreasing numbers.
All of this depends on if the individual is exercising at the lower end of the recommended heart rate or near the maximum. I would guess that getting near the maximum would accentuate the changes.
I hope this helps as you consider heart rate as a time series variable. I am certain others who work with this type of data would have superior conjectures. Hopefully, we will hear from them.
HEART BEAT: One heart beat is defined by a cycle consists of three components: (i) systole, (ii) diastole, and (iii) short pause before the beginning of the next cycle of systole.
DATA: Counting the beat depends on your source of data. Generally, there are two sources: (i) electrocardiography (a print out in a graphical form), and (ii) phonocardiogram (a measurement of the wave made by sound. Electrocardiogram is the most common output used in hospitals and clinics. In electrocardiogram, one beat equals one peak. The distance between one two peaks is considered one cycle. Each peak is a beat. See link below.
TIME SERIES: It is correct to classify this data as time series. In fact, it is a good model to demonstrate "stationality", i.e. differentiate from integrated data. For a short-time observation; it may be used to demonstrate stationarity if a stimulus (shock) is introduced---the heart beat rate will change (going faster) in response to the shock, and after a while will return to its mean. This mean reverting is a good model to demonstrate data stationarity of time series. If the subject is study over a long period, i.e. in months or years, where the shock is a permanent, i.e. raise cholesterol or some kind of permanent or long-term heart affliction where the heart beat rate differs from those collected from prior/earlier period---this is a demonstration of integrate data where the data is not mean reverting.
I am afraid I think it is not possible to do what you want. It seems what you want is to extract the sequence of RR intervals from your heart rate sequence sampled with an interval of 5s. Clearly what you have is an average of the heart rate. It might even be a moving average of the 5 seconds intervals, but even assuming it is not a moving average you cannot extract back the time intervals of the individual heat beats (the RR series).
If you are trying to acquire the heart rate at every 5sec what you first need to do is extract two adjacent peaks. Then you need to find the distance between the two peaks. Divide it by the sampling frequency. U will time for one interval. Multiply it by 60 you will get beats/minutes i.e. the heart rate.
Well, if your HR data were obtained with a monitor that makes the transfer of the HR for RR intervals, just look for this feature in your software. Another way is to save the data column of HR in a txt file and copy them to the kubios software that is available on the internet. This software will transform their values of beats in RR intervals and present the analysis of heart rate variability. Do not apply the analysis by linear methods, because the data that you have were obtained during effort and in this situation this type of analysis is not reliable.
Saba, if your data were recorded with spaces of 5s, you need first to do like Parth Shah told you, ok? Then you can create a column of values and insert them in Kubios, ok?
Kubios as Mario Paschoal sir suggested can be a easy option, Or else if you are good with softwares like LabVIEW or MATLAB then you can use them. If you want any help related to LabVIEW or MATLAB feel free to contact.