Can I create a marginsplot with likert scale variables containing non-integer values

[patricia oreilly]

I am running a regression that tests the interaction between children being read to at age3 and at age 5 on their reading score at age 9. The variables for reading at age 3 and age 5 are scores rather than a dichotomous variable (from likert scales so not proper continuous vars) and contain non-integer values.

Read3yrs	Freq.	Percent	Cum.
0	112	1.62	1.62
1.428571	148	2.14	3.76
2.857143	312	4.51	8.27
4.285714	546	7.89	16.17
5.714286	523	7.56	23.73
7.142857	725	10.48	34.21
8.571429	342	4.95	39.16
10	4,208	60.84	100.00
Total	6,916	100.00

I am trying to produce a marginsplot for interactions between these two variables and gender (female). I have succeeded in producing the margins plot when I treat these two variables as continuous but would like to know A. is this a good thing and B. Is it possible to plot how the reading score at 9 varies with the intensity of reading at 3 or 5 for boys and girls? (so for different values of the read3yrs).

Here is what I have managed to do so far
reg Readscaleage9 mainindependent $controls i.female##c.Read3yr i.female##c.read5yr
est store interaction1

margins, dydx(Read3yr read5yr) at(female=(0 1 )) vsquish
marginsplot