sample section of "getting started" doc

[schluppeck]

hi pierre,

i had a bit of time to put together the first few paragraphs of a getting started page. a couple of things:

1. it's probably slightly verbose, but probably better to err on that side for newcomers
2. wasn't sure whether you'd prefer a separate folder or if files in there are ok. can always reorganise

the next few sections could/would add more complexity (in order of what people think is most important to explain).

happy to brainstorm about how to make this work best... denis

[piermorel]

Hi Denis,

Thank you, it seems great to me. I don't mind the verbosity, which is needed for people who are not familiar to OO Matlab. I just hope the psychophysics data won't be too disconcerting to people outside the field! On that I was wondering about the possibility make the data more explicit (using separate variables, a structure or a table rather than a matrix without named columns). One cool thing about psychophysics data is that a more advanced section could have the raw responses rather than averaged ones and use stat_ functions to generate the same plot but from the raw data ( stat_summary() with 'type','fitbinomialci' and/or stat_glm() ).

As for the general organization of the help, we have to see: a folder might be nice if there are many files. One possibility would also be to use Matlab's publishing features instead of a .md file, similar to what I use for examples.m and the generated /html folder. This way you get runnable code and a formatted output that is easy to open for everyone (I'm not sure how far back in Matlab versions that would work though).

Oh, and gramm now has an export() method that directly creates an svg file ;)

[schluppeck]

ok - all sounds good!

• re data sets: could stick to standard matlab data sets for the first few figs in the tutorial and introduce something like the psychophysics data later (i guess gramm depends on stats toolbox, so all of the datasets included along it would be ok https://uk.mathworks.com/help/stats/_bq9uxn4.html ) - something like fisheriris.mat would also be recognizable by non-neuroscience people. but I don't think a neuro/psychology dataset will put people off ... as long as it's explained.

• i also thought about the stat_summary() option for doing the fitting "properly" with that psychophysics dataset - i think that's a great idea that shows how useful the approach is... that's definitely on the list for showing how to build up layers.

• can stick to Matlab publish() - which makes things more self-contained and avoids copy+paste from markdown. That feature has been in Matlab for a long time (not as a GUI button, but definitely as publish.m) so should be pretty safe.

will ping you when I've had a chance to do a bit more of this ttys denis

PS: the export() feature is very nice - I like how clean the plots are when exported. I used to regularly fight through EPS/PDF export -> Adobe Illustrator