track-and-graph
track-and-graph copied to clipboard
SamAmco
[Feature request] Plot a series against another
Hello,
Thank you very much for this app! One thing I would like to eventually extract from all the data I'm capturing in it would be potential correlations between series, e.g. level of stress vs. hours of sleep or hours of work, this kind of things. The simple approach would be a cloud of points, which may (or may not) give some useful insights. A fancier approach could add a heatmap-like color scale or variable dot sizes to reflect the frequencies of occurrence of answers. The super fancy approach could use some machine learning algorithms to find correlations between any existing series :) But I'd rather export the data from the app and process it directly on a computer in that case.
What do you think? Cheers, Simon
Hi sorry I do not really understand the problem you are trying to solve exactly? How would a cloud of points or a heat map help you to visualise a correlation? Perhaps you could give a visual example?
Sure. While making the example I realize that it needs some kind of data correspondence, e.g. common dates for both series.
Say, for example, that both series "hours of work" and "hours of sleep" have one point per day, both time series would look like the top graph in the app. What I'm suggesting is to allow plotting "hours of works" against "hours of sleep", e.g. "hours of sleep" on the X axis, which may (or may not) suggest e.g. that I never sleep more than 7 hours when I work more than 8 hours a day:
Adding a way to represent the occurrence frequency would e.g. mark the center dots in dark red (high frequency) while the rare extremes would be in dark blue (very low frequency), or maybe a trendline. I realize that it's not trivial though.
Sorry for the delayed response. This is an interesting idea. The top graph you have there looks like something you could do in track and graph already. The second graph is of course not currently available and as you say not particularly trivial. I do appreciate that there is a large need for better graphs when it comes to analysing correlations between data so this is high on my todo list, but i have very little time for the app, so i often end up taking on smaller issues instead.
I understand, no worries! Yes the top graph is similar to what I already have in the app. In any case, the data is always easily available for further analysis with specialized tools, so I don't think you should make this a top priority.
On 17 August 2021 22:12:17 CEST, Samuel Harvey-Collins @.***> wrote:
Sorry for the delayed response. This is an interesting idea. The top graph you have there looks like something you could do in track and graph already. The second graph is of course not currently available and as you say not particularly trivial. I do appreciate that there is a large need for better graphs when it comes to analysing correlations between data so this is high on my todo list, but i have very little time for the app, so i often end up taking on smaller issues instead.
-- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: https://github.com/SamAmco/track-and-graph/issues/115#issuecomment-900598161
Just came here to say I love the app, and this would be my number one feature request. (More ways to analyze correlations).
Even just starting with a scatter plot and showing the Pearson correlation coefficient would be really helpful for identifying correlations between two Trackers.
@SamAmco Request to take a look at this library https://github.com/PhilJay/MPAndroidChart & see if we could pick some graphs for our usecase.
Hi. I am aware of that library. The library T&G is currently using is androidplot. I don't think there's a major obstacle to using androidplot for stuff like this so I would rather not have 2 graph libraries. However this specific graph type just isn't high on my priority list right now i'm afraid. The data stored by track & graph is just a sqlite database, you can get it using the backup feature and saving it to a file. You can also export your data to CSV, so if you need more powerful visualisation tools you might want to take your data and put it into something else.