Problem in plotting interatcion term after fractional logit regression

[Lydia Lu]

Hi, Everyone:

I have two questions need your help.

(1) I use fractional logit regression and have an interaction term. May I know how to plot the interaction term after running the fractional logit model by STATA?
because it always shows: factor variables may not contain noninteger values, when I use "margins" command.

(2) Previously, I use excel to plot the interaction term by using mean value adding and subtracting one standard deviation to represent the high and low-level moderator.
My regression formula is Y=a+bx+cm+dx*m where the coefficient for margin effects on b is 0.003, c is 0.077, and d is -0.002,
the mean value of independent variable(b) is 6.04, s.d=8.62, the mean value of moderator(m) is 0.56, s.d=6.25.
the problem is when I use excel to plot interaction term, it generates negative Y value. However, my Y value should be range from 0 to 1,
may I know what happened in my model and how do I solve this problem?

Very thanks for your help.

[Richard Williams]

Show the fracreg and margins commands you entered. My guess is that you are not using factor variable notation right, but without actually seeing the commands I can't say.

Also, I like using the user-written mcp command for the type of plotting I think you want to see. For an overview, see

https://www3.nd.edu/~rwilliam/stats3/Margins03.pdf

This code with fragreg works fine.

 webuse 401k, clear
fracreg probit prate mrate c.ltotemp##c.ltotemp c.age##c.age i.sole
margins sole
mcp ltotemp sole, var1(20)

[Lydia Lu]

Dear Williams:
Thank you so much, and I list my command in the below
fracreg logit eacountdivttlcitation_add0 roapasup_1 rever_roapasfall_1 wcpsales_1 rever_roafallwcpnew_1 logfirm
> age_1 logsales_1 zscore_1 inventratio_1 der_1 perrd_div100_1 patcount_add0_1 instiowner_1 ipc0_fhhi1_1 fselfr
> ate0_1 bselfrate0_1 indsample_1 indhhi_1 log_indsalesgrowth_1 if year>=1998 & year<=2012, nolog


Fractional logistic regression Number of obs = 2,436
Wald chi2(18) = 774.98
Prob > chi2 = 0.0000
Log pseudolikelihood = -1065.0537 Pseudo R2 = 0.2982

--------------------------------------------------------------------------------------------
| Robust
eacountdivttlcitation_add0 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
roapasup_1 | -.0252506 .0143891 -1.75 0.079 -.0534528 .0029516
rever_roapasfall_1 | .0146206 .0071983 2.03 0.042 .0005122 .028729
wcpsales_1 | .5758466 .171755 3.35 0.001 .239213 .9124803
rever_roafallwcpnew_1 | -.012656 .0039248 -3.22 0.001 -.0203484 -.0049636
logfirmage_1 | -.820365 .2143916 -3.83 0.000 -1.240565 -.4001652
logsales_1 | 1.579781 .108204 14.60 0.000 1.367705 1.791857
zscore_1 | .0032741 .0089834 0.36 0.716 -.014333 .0208812
inventratio_1 | -.412029 .664924 -0.62 0.535 -1.715256 .8911981
der_1 | -.0047975 .0013124 -3.66 0.000 -.0073696 -.0022253
perrd_div100_1 | .0637623 .0128751 4.95 0.000 .0385276 .0889969
patcount_add0_1 | -.004918 .0006848 -7.18 0.000 -.0062601 -.0035759
instiowner_1 | -.0115741 .0027754 -4.17 0.000 -.0170139 -.0061344
ipc0_fhhi1_1 | 2.49512 .1854303 13.46 0.000 2.131684 2.858557
fselfrate0_1 | 2.080399 .2203042 9.44 0.000 1.64861 2.512187
bselfrate0_1 | .8095915 .7241519 1.12 0.264 -.6097201 2.228903
indsample_1 | -.0075336 .003381 -2.23 0.026 -.0141602 -.0009069
indhhi_1 | .0027345 .0079212 0.35 0.730 -.0127907 .0182597
log_indsalesgrowth_1 | .234604 .1180456 1.99 0.047 .0032389 .4659691
_cons | -10.93776 .7752999 -14.11 0.000 -12.45732 -9.418197
--------------------------------------------------------------------------------------------

margins rever_roapasfall_1#wcpsales_1
rever_roapasfall_1: factor variables may not contain noninteger values
r(452);

I will quicly try your suggesitons.
Very thanks for your help.

[Richard Williams]

Hi Lydia. First off, your code and output would be much easier to read if you used code tags. See pt 12 of the FAQ.

Second, as I suspected, you aren't using factor variable notation. You may have even computed dummies yourself, which is another no-no if you are going to use margins. Or, the two variables you are trying to use in the margins command really are continuous, in which case it is not legitimate for them to appear to the left of the comma.

Why don't you take a look at

https://www3.nd.edu/~rwilliam/stats3/Margins01.pdf

If you like that, you may want to check out the other margins handouts at https://www3.nd.edu/~rwilliam/stats3/index.html

If it is still unclear to you what you are doing wrong, write back.

[Lydia Lu]

Dear Williams:
Thanks for your great material and I have read it and also install mcp commands.
However, I still have problems to plot "interaction" term, would you mind if I can send
my data and my commands to your email and request your assistance?
Best Regards
Lydia