Ask HN: How to dive in to basic applied statistics?
The tool that does not seem to exist should do this:
- choose what you roughly want to display (e.g. values over time)
- throw some data at it
- it auto-detects file format, dates, values, table headers
- it shows a live parsing preview, maybe even a rough plot
- you can tweak the selection, but more interactively than with regex and with a live preview by highlighting within a few example rows of your data
- plot some nice, live tweakable graphs, starting from presets
- guide through some basic transformations to do (like filtering)
- enable to update the source data
- export images or html
- for more exessive logs I'd whish to have quick filtering. If I log and tag every click of a user on my website I might for example want to select behaviours within the first week of registering (account created at XY, active within XY etc, clicks on the settings menu, sort by most clicked and display "button label")
The guide I want:
- what can I plot fast to gain some insight, e.g. for my business?
- where do I have to take care of pitfalls and should not just plot ahead (biases etc)?
- how to extract correlations? (and which are actually meaningful?)
- how can I build a simple model of something and see how changes in input change the output?
I know the guide part is just not that easy and one has to keep a sharp mind not to be misguided, but one has to start somewhere and even basic statistics can be right and really helpful (churn-rates etc).