This is an app I made for creating time series datasets using a visual interface: www.dspencil.com
I had the idea because quite often I find myself sketching out time series plots to explain some concept, so I thought that it would be nice to do this and then get an actual dataset out of it. You can:
- Draw variables freehand by clicking on the graph.
- Add random noise variables.
- Add correlated variables with a specified correlation coefficient.
- Interpolate your variables to fill nulls.
- Download your dataset as CSV or feather.
- Share your dataset, this saves it to a database and gives you a unique code to load it with.
- Currently supports daily, monthly or weekly frequency data.
Would love to hear any feedback!
Hi @robert_d Robert,
Welcome to the community. And thank you for the shoutout.
I think your app is awesome and unique. I can’t recall seeing anything similar to your dspencil app.
Filling the variable nulls with cubic interpolation
@robert_d This is intriguing. if you had a video to go with this it could help showcase use cases (and limits)
Thanks for the comments @adamschroeder and @DaveG! A how-to-use video is a good idea, when I can find some time I’ll make one.
As suggested, I’ve added a short how-to-use video: https://youtu.be/3aeLCcHTnyY
perfect … thx @robert_d … covered everything nicely … the share option is a nice add-on … and the multiple interpolation options … I wonder how granular a time series / how many points this might be used for
would you see adding two variables (in the app) … e.g. adding a base variable w a noise variable … as presumably could do that outside of the app as well (presuming a 1:1 correlation of time range and intervals)
Hi @DaveG, an option to add variables together is a great suggestion. I had some time so I’ve added that as a feature: ‘add combined variable’. Also added ‘add shifted variable’, which is basically just the pandas ‘shift’ method.
Thanks for the feedback and ideas