THIS PROJECT IS VULNERABLE TO CVE-2021-24112 ON LINUX AND MAC-OS. THIS PROJECT IS NOT BEING MAINTAINED, USE AT YOUR OWN RISK
This vulnerability is labeled Critical with a CVSS score of 9.8/10
A currently very basic CLI for some stats stuff. It can list and it can summarize a numerical dataset. For 2D data it may also plot it. Keep in mind, that regardless of the number of dimensions, the last item in an ordered pair is taken to be the coordinate of the RESPONDING variable.
wstat set=1,2,3 list
Will print something like this
1.000000000
2.000000000
3.000000000
You can also do wstat set=1,2,3 operation=list if you're not a fan of the shorthand. The command line arguments can go in any order. For example wstat list set=1,2,3 is equally valid. In addition, list is the default opperation, so wstat set=1,2,3 is enough.
You can also use scientific notation in your set, for example wstat set=1.1e-2,1.0e-2,0.9 would work fine. You can use a capital or lower-case "E".
Lists are cool and all, but at the end of the day, what you care about is probably more down here. wstat set=1,2,3 summary will print something like this:
These are all rounded values. If you need more precision, use JSON output
Min Q1 Med Q3 Max
1.000000000000 1.000000000000 2.000000000000 3.000000000000 3.000000000000
N (Set Size) Mean Std. Dev. (s) Std. Dev. (σ)
3 2.000000000000 1.000000000000 0.816496580928
Possible Outliers:
This works just as well in N-Dimensional sets.
You can also use this summary command to indicate potential outliers. They will be right under your normal summary.
Type correlate or correlation to print the r correlation coefficient
It's not terribly convenient to pass in everything as a command line argument, so you can pass in your set into stdin. This means that if rng is a command that creates a comma-separated list of numbers, rng | wstat summary will let you summarize that list easily.
In addition, you can also give no input and no command-line parameters: wstat summary and the console will let you write your input, which will be read when you hit enter. This is useful when you are copy-pasting your dataset from elsewhere.
Currently json, csv and text output are supported. Text is the default. You can specify with wstat list json or wstat list text. The full version is wstat operation=list output=json. The same information is given in both the list and the summary, it's just displayed differently. Please note that the only operation that supports CSV output is List.
You can specify an output file with wstat -o C:\Users\benny\Desktop\a.txt or wstat file=C:\Users\benny\Desktop\a.txt The directory you point to must exist. You can read a set from an input file with wstat set=C:\Users\benny\Desktop\a.txt or specify the file with stdin.
When a file path is specified as a commandline argument, if it has spaces, it must be in quotes.
This supports n-dimensions, just add dimensions=2 to your command. A shorthand in the case of 2-D data, is 2var. If you have 3+ dimensional data, you need to use dimensions=3
Pretty simple, just do wstat 2var plot and once you type in your data it will give you the filepath.
For example, a set like this: (-3,9),(-2,4),(-1,1),(0,0),(1,1),(2,4),(3,9)
Would produce a plot like this:
The parantheses around points are optional, I include them because I prefer the notation, especially for a demo/tutorial.
As of Alpha 7.0.0, linear regression is supported on multidimensional (2+ Dimensioned) sets. Just use wstat dimensions=3 linreg for 3 or more dimensions.
Add a linreg to your plot command wstat 2var plot linreg and it will give you a printout as well as draw the Least Square Regression Line on the graph. linreg is a shorthand for options=linreg, if you use multiple options you can do options=option1,linreg,option3 or you can use the shorthands together.
You can also use linreg alone, such as wstat 2var linreg, which will do the same thing but without making a fancy picture. This is ideal for CLI-only OSes like Ubuntu Server, or when you quite frankly don't care about the pretty pictures. It is also required for 3 or higher dimensions, as you might imagine it is difficult to display a 77-Dimensional plot.
For the set -2,2,2,0.11134,0,0.8 I got this plot:
And this printout:
y=b0 + b1x1 + b2x2 + ... + bnxn
b0 = 0.9704466666666667
b1 = -0.472165
Coefficient of Determination (r^2) = 0.9761489873568633
Filepath: C:\Users\benny\source\repos\stat\statistics-cli/plots/plot_2019-10-28___08-35-24_PM.bmp
Note that this printout is also supported for JSON and CSV output. CSV output does not contain the Coefficient of Determination.
Add a reexpress, and then choose either zscore or residual as an option. zscore and residual works on N-Dimensional sets. residual also requires you specify the regression line to use.
wstat reexpress zscore will standardize a 1-Dimensional set.
wstat reexpress zscore 2var for 2 dimensions.
wstat reexpress zscore dimensions=<n> for n dimensions.
wstat reexpress residual 2var linreg will print the residual set for a linear regression line in 2 dimensions.
wstat reexpress residual dimensions=<n> linreg will print the residual set for a linear regression line in n dimensions (Since Alpha 7.0.0)
Type wstat cdf for the cumulative distribution function for a normal distribution. Type wstat invCdf for the inverse.
Type wstat quantile to split the set into quantiles. Choose the rank of the quantile with rank=<n>. Default rank is 100 (percentiles)
For drawing plots you will need two packages: libc6-dev and libgdiplus For some distributions, xdg-utils may not come preinstalled, and that is needed for auto-opening images. If you are not using this app for drawing plots, you do not need to install these packages.
For APT
sudo apt install libc6-dev
sudo apt install libgdiplus
xdg-utils (Probably already installed)
sudo apt install xdg-utils
For yum
sudo yum install glibc-devel
sudo yum install glibc-devel.i686
sudo yum install libgdiplus
xdg-utils (Probably already installed)
sudo yum install xdg-utils
It should work, but I have no idea how well it will work, I don't have a Mac to test it on, and I am hesitant to pirate OSX and stick it in a VM. Let me know how it goes if you try it.
I have only built x86-64 binaries, but as it is open source, anything that .NET Core will compile on should work. The reason for this, is that I cannot test on a processor that I do not have.
Keep in mind, required packages may differ if you run on 32-bit x86 or on ARM.
This project is a very low priority for me, so if you decide to open some pull-requests to add features or otherwise improve code quality I would be very thankful.
Please note, I would like to have CI sorted out on PRs, however github actions is not currently cooperating with .NET Core 3.0, so we are currently stuck without it.
In addition, everything is subject to code-review
Licenses of all dependencies are included in /Dependency Licenses/, any derivative works of this project are to comply with those licenses in addition to the license of this project which is included in /LICENSE
