Likert Scale Graph with Mean Score

[winkhaing]

Could someone suggest me how

to draw Likert Scale Graph with Mean Score like below?
Thanks.

[Joseph Coveney]

Sorry for the ignorance, but:

What do the stacked bars represent? Proportion of responses in each category (e.g., brown=strongly disagree, peach=disagree somewhat, blue-gray=neither agree nor disagree, light blue=agree somewhat, blue=strongly agree)?

Why are they offset to different positions along the x-axis? Nonresponse?

[Nick Cox]

It looks as if the graphs are centred (centered) on the middle category. slideplot from SSC, a command from 2003. does almost everything that graph does, except for that centring. Writing an alternative using graph twoway has been on my mind for about 5 years, but usually I use tabplot from the Stata Journal instead.

In Stata, some mix of twoway bar and twoway rbar would get you what you want.

[ericmelse]

As was discussed before. Several solutions are discussed here and here. Also Ben Jann's paper 'Tabulation of multiple responses', that was published in The Stata Journal (2005 Vol. 5 ,1, 92-122) offers solutions to your 'check box question' and is downloadable here. You should also consider coefplot, discussed in another paper from Ben Jann, which is downloadable here.

[Nick Cox]

Eric Melse Excellent links. I'd forgotten that earlier work.

[Joseph Coveney]

It looks as if the graphs are centred (centered) on the middle category.

Thanks, Nick. I hadn't noticed that. I learned from the links that Eric gives that it even has a name, "diverging stacked bar chart".

[Nick Cox]

That's an old idea given a strong boost by some recent advocates. I try to maintain an open-minded attitude of encouraging whatever works best for your data and your purposes

But now the divergers have diverged. See https://www.datarevelations.com/reso...kingdivergent/ for an argument that strongly disagree and strongly agree should be next to each other!

[ericmelse]

For those users who visit this post and might be unaware, note that Nick published the module floatplot (29 May 2021, but now in version 1.0.3, 31 October 2022):

to draw floating or sliding stacked bar plot - floatplot produces a floating or sliding stacked bar plot showing percents (or optionally proportions or frequencies) of categories of a numeric outcome variable numvar by zero, one, or two other categorical variables. The plot is most helpful if categories of an outcome variable have a natural or conventional pre-defined order. Although there is no formal check, the design of the plot tacitly assumes a modest number of distinct categories, say between 2 and 9. The commmand is a wrapper for a call to twoway rbar for showing bars and twoway scatter for showing text.

It can be installed by using this code:

ssc install floatplot , replace

The examples in the help demonstrate how to create a floating or sliding stacked bar plot for frequencies, proportions, or percents. Very useful indeed to visualize the so-called Likert scale response variables also over categories.
Note that the above is also discussed in this post.