1
$\begingroup$

I am working on a dataset with physical measurements taken daily (weight, bmi, etc...) and I am working through the process to graphically represent it. I think it is worth noting that every day has a corresponding row, but if no measurements were taken, the values are the same as the day before.

Here is an example of the trend I am trying to manage:

Date, Weight, BMI
1/1/2016, 155.1, 21.9
1/2/2016, 155.1, 21.9
1/3/2016, 155.1, 21.9
--continued for several weeks--
3/1/2016, 170.2, 25.0
3/2/2016, 170.1, 25.0

edit: I should clarify that these repeated values are how the data is put together for missing values. The days that are repeated are days when no measurement was taken

If the value stays the same for an extended period, should there be a gap in any graphical representation? Should I keep the numbers as they were (155.1 & 21.9) until the next measurement was taken, or should the numbers increase over that time to "bridge the gap" - meaning they would increase by the difference in measurements divided by the number of days?

It feels like I should be increasing the value over time to account for what would happen in reality, but I don't know if this would have negative implications on the data.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

I'd represent actual repeated values differently from missing data. The latter is straightforward; the former I'd interpolate with Gaussian process regression. This way you'll be able to get error bars like so:

Gaussian process regression

Note how the sample functions (and thus error bars) expand and contract as you depart and approach the measurements, as you would intuitively expect.

Improve this answer
$\endgroup$
2
  • $\begingroup$ I have added an edit to the post. Does your answer change knowing that the repeated data points are actually gaps in the data? The dataset is put together this way, but I know that no measurements were taken on those days $\endgroup$
    – Lbutlr
    Dec 14, 2016 at 18:34
  • $\begingroup$ I'd consider all the repeated values missing data and treat them as outlined in my answer. If I were to visualize it I'd make it clear the data is missing instead of using the last measurement. $\endgroup$
    – Emre
    Dec 14, 2016 at 18:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.