How to recenter the Mediator?

[Andyzr]

Thanks for your wonderful work!

On page 15 of the paper Acharya, Blackwell, and Sen (2016), there are illustrations about "recentering" the mediator to visualize the relationship between ACDE and different controlled values of the mediator:

In addition, we could investigate the causal interaction side of this analysis and find the ACDE(m) for
different levels, m, of current-dayincome. We do this by simply recentering M_i to different values and rerunning
the sequential g-estimation

I try to mimic the recentering process by the following code:

summary(direct <- sequential_g(women_politics ~ plow + agricultural_suitability + tropical_climate +
  large_animals + political_hierarchies + economic_complexity + rugged |
  years_civil_conflict + years_interstate_conflict  + oil_pc + european_descent +
  communist_dummy + polity2_2000 + serv_va_gdp2000 |
  I(centered_ln_inc+2) , data = ploughs))

However, the coefficient of plow does not change, while it should be changed to positive as indicated by Figure 8 in the paper Acharya, Blackwell, and Sen (2016) :

t test of coefficients: 

                            Estimate   Std. Err. t value Pr(>|t|)  
(Intercept)               9.67870909  3.96785647  2.4393  0.01626 *
plow                     -4.48745622  2.29418073 -1.9560  0.05291 .
agricultural_suitability  4.99910241  2.74485497  1.8213  0.07119 .
tropical_climate         -3.32411998  2.14462889 -1.5500  0.12392  
large_animals            -1.30890824  3.81781460 -0.3428  0.73235  
political_hierarchies     0.43275754  1.01806758  0.4251  0.67158  
economic_complexity      -0.00092324  0.41412996 -0.0022  0.99823  
rugged                   -0.28208461  0.49412866 -0.5709  0.56921  
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

I thought I am missing something. Could you please elaborate on how to recenter the mediator?