Wednesday, February 29, 2012

Carta Terunggul


Sempat menghadiri satu sesi menarik ttg thesis hari ni. Cumanya, hati agak gusar sbb penceramah tu berkata, kalo byk bar chart dlm tesis itu bkn lah tesis yg bagus, kerana bar chart adalah utk budak 12 tahun. Dia actually pernah bg comment mcm tu dkt pelajar PhD yg dia examine, dan dia ckp, mmg student tu sgt sentap dgn dia (penceramah ni background pure sciences, if anyone needs to know).


Belek2 balik kerja, hmm.. bole tahan byk nya jg bar/column chart yg aku ada so far ni. So, terpaksa tangguh kjp kerja at hand, and study balik jenis2 chart and how to create more professional-looking graphs. Blog entry pon sama, habis hilang mood nak letak gmbr2 happy yg telah di pelan, sbb, well.. I just don't feel that happy sgt nih hah...


Anyway, for my own notes, and kot bole tolong sesapa, here's a gist of what I gathered today. I put the links instead of copy pasting the diagrams on their web here, because I had learnt from my GTA class that it is wrong to copy (it's plagiarising) diagram/photos sesuka hati even kalo dah bg kredit pon in the end, kalo imej itu tiada Creative Commons licence. Oleh sbb malas nak bg kredit / cek status imej sesatu, I buh links je lah ye. Anyway, from the keywords pon dah cukup utk Google dan sendiri teroka, Insya Allah:


1. For long labels, change orientation of bar graph - read more here


2. Sort labels by values, not arbitrarily or alphabetically - read more here


3. Try using single clustered bar chart - read more here


4. Alternatively, can try using dot plots instead of bar chart - read more here


5. Even better, try using multi-panel dot chart - read more here


6. Start bar chart from zero baseline to avoid misguidance of data - read more here


6. Remove non-white pixels on graphs - read more here


7. For line graphs, never connect the dots, but give best-fit line instead - read more here



Those yg kalo mmg byk bar chart, I'm not saying it's wrong, heck, I have quite a substantial amount of them, coz I thought they would make the examiners' lives a lot easier kalo senang nak interpret, but I'm going to have to discuss this with my SV first and see if he can give me an HONEST opinion of what he thinks of my 'primitive' charts. So far dia OK je, but I'll discuss with him later once I have done some adjustments to them, to see if he thinks they are better after being 'fixed' and 'fancifed' (the latter is not a word, btw).


Harap2 aku bkn sengaja mencari keja extra x pasal.


4 comments:

Lia Sakinah said...

Thanks noris for sharing this wonderful info. Never come across my mind that a bar chart is more suitable for kids! huhuhu...

mast@work said...

aaaa kenapa la takde button favourite.. ngahaha.. tq for the info.. bagus ni..

mas mmg pernah terbaca somewhere too many bar chart is not good.. mcm kurang mantap gitu

sumpit said...

group aku suka sgt boxplot, sbb it shows a lot. Smpai sekarang i still cant get the hang of it.

Aku rasa ramai pon tak familiar dgn boxplot. Hari tu sorg reviewer paper aku comment tah pape about the graph, which we can conclude yg sebenarnya dia tak paham boxplots.

Maybe u shd consider boxplots too ;)

Oyis said...

lia: tu la kan. tp kalo measurement kita byk buat against time (like simulation mcm lia buat), x mungkin end up dgn bar chart kot. at the very least line graph kan?

mast: oo kurg mantap ye? tu la, skg tgh tgk2 balik sesatu graf ni, kaji balik keperluan graf tersebut, or if ada alternatif lain from bar chart, guna. i think lps diskas ngan sv i kan, masalah dia bkn dkt penggunaan bar chart sgt, tp bar chart yg terlalu simple dan seketul2 (jd byk ketulan simple). kalo buat clustrered bar chart not so bad kot.

eti: masya Allah, boxplot, mcm mana nak interpret tu? (aku tgk contoh dia). baru jenguk2 potensi menggunakan R dah pening2, tgk boxplot leh pitam terus aku. aku rasa at this stage, i'll stick to what i know lah. nnt balik msia aku blajo ngan ko eh, leh cuba nnt wat paper, insya Allah