0
$\begingroup$

I hope it is permitted to ask this type of question here.

I asked many participants to grade the fidelity (i.e resolution) of different 3D models of different resolutions on a scale of 1 star to 5 stars (The Scale:"Bad Resolution" = 1 star, "Poor Resolution" = 2 stars, "Fair Resolution" = 3 stars, "Good Resolution" = 4 stars, and "Excellent Resolution" = 5 stars ). I have obtained the dataset (shown in image - kindly check) enter image description here

PS: results are fictitious in order not to breach ethics.

Number of participants is quite high so a 3D model variation like HeadofRamesses50K will be graded on the scale of 1 to 5 by many users.

Are there any type of charts that might show correlations in such situation? Like how many percentage of participants' evaluation each model variation was graded on the 1 to 5 scale shown in a nice way? Any concrete examples in R/Gnuplot would be appreciated.

Improve this question
$\endgroup$
9
  • $\begingroup$ On the face of it (pun intended) you have different model names and the numbers of ratings 1 to 5. What is practical depends on how many distinct model names there are. In your example I see just 9, so a 9 x 5 bar chart of some flavour should work fine. $\endgroup$
    – Nick Cox
    Mar 14, 2017 at 16:12
  • $\begingroup$ Welcome to Cross Validated! Please take a moment to view our tour. $\endgroup$
    – Tavrock
    Mar 14, 2017 at 16:56
  • $\begingroup$ If you want to show the relative dispersion of each of the sets of answers from your 1000 or so responses, a row of boxplots may be effective. You could also display boxplots by resolution rating, to see if that has overall consistancy. $\endgroup$
    – Tavrock
    Mar 14, 2017 at 16:58
  • $\begingroup$ @Tavrock Boxplots really don't work well when the raw data are 1 2 3 4 5. Relative frequencies can be very hard to read off. $\endgroup$
    – Nick Cox
    Mar 14, 2017 at 19:34
  • $\begingroup$ @Nick Thank you so much for your comments and suggestions. Actually I have 17 variation of 3D models (i.e 17 different resolutions of 2 models) but I did not put them in the example for brevity. So it would be 17 *5. Not to mention 34 participants :-) I will try to see if a barchart in R would solve this and get back to you. Thanks again. $\endgroup$
    – HB87
    Mar 14, 2017 at 19:39

1 Answer 1

Reset to default
0
$\begingroup$

Thanks for the suggestions. The following is what I reached so far. I wanted to create Nick Cox's graph in R- the tabplot shown Nick's tabplot in Stata code.

I am not successful till now. Any suggestions on how to do that type of plot in R or Gnuplot would be greatly appreciated.

I managed to create a Mosaic plot but still need a lot of tweaking (most important a legend). I have 2 models with 17 variations. I thought I could fit them all in the same graph but the labels at top of the upper "axis" did not fit even with being perpendicular to the axis (las = 2). To solve this I used plyr package (the revalue function) to shorten names of model versions. The following graph is for one of the models which has 7 variations of resolutions (7*5).

enter image description here

My R code so far is as follows:

mosaicplot(table(theMercuryDatawithNoInt), color = TRUE, main="", xlab = "Model Versions", ylab = "Rating with No Interaction", legend=TRUE,  cex.axis = 0.66, las = 1,direction = "v")

One crucial thing here is there is a lack of space for the legend. Also I don't like where the axis labels are. In addition, I am receiving the following warning as some arguments like the legend being disregarded for lack of space maybe:

Warning message: In mosaicplot.default(table(theMercuryDatawithNoInt), color = TRUE, main = "", xlab = "Model Versions", ylab = "Rating with No Interaction", legend = TRUE, cex.axis = 0.66, las = 1, direction = "v") : extra arguments ‘legend’, ‘direction’ will be disregarded

Any suggestions on how to fix this. Other better answers or other graphs are also appreciated!

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.