Data analysis software

bump, also interested but have nothing to offer

Python + libraries
R + packages
Julia
Matlab if you're a retard or work with retards

Big talk from someone who doesnt have the ability to understand, much less reinvent the wheel.

matplotlib is pretty good if you don't want to bother with gnuplot scripts

Paths of Glory?

make your own software faggot

>Matlab if you're a retard or work with retards
The irony

Depends on the data really, Origin offers a nice GUI for analysing DSF data and producing publication quality graphs, however it's click intensive. I'd prefer if I could do it in R, which I probably could but I've never learnt.

For basic stats I personally would use R because I learnt that way, however I suspect python is probably the easier way. Numpy + Matplotlib + SciPi + Pandas have great functionality, compared to the many more packages you may need to install if yo use R.

Don't use matlab.

How is pic related for someone new to stats and programming in general?

Attached: 51XOP1zda3L._SX376_BO1,204,203,200_.jpg (378x499, 34K)

R, Mathematica, Excel

Other languages confuse me because they are not functional...

I've never read that book, but books that combine introductory stats with stats programming typically don't go deeper into statistics than they need to for introductory stats programming. Using a intro stats textbook alongside it might be helpful for a fuller understanding of the statistics concepts, as it will have more detailed descriptions and more practice problems you can do to check your understanding.

other books from that publisher aren't terrible

Gnuplot is great for testing purposes. Its fitting capabilities are great and it's very fast, but getting a quality output is quite difficult and tedious. Also integration and automation isn't really good.

Python + matplotlib is great, can produce publication quality output easily, and may be integrated into larger programs. It's quite scripty though, and I often find myself writing lots of lines of code for a simple graph because of all the necessary data processing and customisation.

Julia = Perl + Python + Javascript + C + R + Lisp

I sort of want to get started on the Julia train.
But the emacs setup could be better and man do I hate the emoji programming and the whole "muh Greek letters in source code :DDD"

I have a stat background and used to think R vs Python was a valid comparision, having used both extensively I know think that they have fairly complementary strengths. Essentially if your data is text or images python is better, if your data is numerical/categorical, especially if organized as a table R is preferable.

The deep learning toolkits almost all have bindings to python while they generally don't have bindings to R, the fields deep learning excels is in CV and NLP, and python is used widely there additionally there isn't really anything like PIL in R. Comp Imaging still uses MATLAB quite a bit, so if you don't care about deep learning MATLAB would probably suit your needs for that sort of image analysis.

I think where R really excels is when doing any sort of linear modeling or statistical analysis, if you care about understanding relationships between variables the tools R has in its standard library are way better than anything scipy has and CRAN has tons of packages for more niche analyses. Additionally R far surpasses python and MATLAB when it comes to running simulations.

Another argument I see is that python has fewer but higher quality packages while R has many but variable quality packages. I think people have this idea by comparing numpy/scipy to the average CRAN package. Much of CRAN are packages written for highly specific use cases or as part of active research being done by a lab, while numpy/scipy is essentially rewrite of the MATLAB stdlib + some toolkits, and pandas an implementation of R data.frames. These all end up clunky but serviceable versions of their originals, that are required to do things that R/MATLAB do innately. R does have a set of highly useful packages dplyr for data manipulation and ggplot2 for plotting spring to mind.

I think OOP holds python back from data analysis to some extent. It also maybe shouldn't be surprising that deep learning is being done in python as there actually are some benefits to using OOP there.

Depends on the data, but python works for everything and R is super elegant in certain cases.

Matlab is pretty good even though it can be awkward and inconsistent in its conventions sometimes.

depends on the tasks
but Origin is in general really fucking good