sim_slopes seemingly not incorporating weights from lm() objects

[tgravelle]

Hello, in probing an interaction from a weighted linear model fit using lm() the simple slopes analysis from sim_slopes() that is output does not (apparently) make use of the observation weights.

For example, the model and simple slopes analysis below...

m_wtd <- lm(FB_manipulate ~ fb_fact_check
            + Gender + ln_age_years + Edu + RaceEthn + Region + Consp_Mentality + Ideology + PID
            + fb_fact_check*Consp_Mentality*Ideology, data = data.2a, weights = weight)

sim_slopes(m_wtd, pred = fb_fact_check, modx = Consp_Mentality, mod2 = Ideology,
           modx.values = cm_min_mean_max,
           modx.labels = c("Low", "Mean", "High"),
           mod2.values = ideol_lib_mod_con,
           mod2.labels = c("Very liberal", "Moderate", "Very conservative"),
           pvals = TRUE, interval = TRUE, digits = 5)

... returns the following results:

███████████████████████████████████ While Ideology (2nd moderator) =  0.509530222 (Very conservative) ███████████████████████████████████ 

SIMPLE SLOPES ANALYSIS

Slope of fb_fact_check when Consp_Mentality =  0.3555157 (High): 

     Est.      S.E.    t val.         p
--------- --------- --------- ---------
  0.07079   0.03593   1.97046   0.04890

Slope of fb_fact_check when Consp_Mentality =  0.0000000 (Mean): 

      Est.      S.E.     t val.         p
---------- --------- ---------- ---------
  -0.01930   0.02311   -0.83507   0.40376

Slope of fb_fact_check when Consp_Mentality = -0.6444843 (Low): 

      Est.      S.E.     t val.         p
---------- --------- ---------- ---------
  -0.18261   0.07417   -2.46222   0.01388

████████████████████████████████████████ While Ideology (2nd moderator) =  0.009530222 (Moderate) ███████████████████████████████████████ 

SIMPLE SLOPES ANALYSIS

Slope of fb_fact_check when Consp_Mentality =  0.3555157 (High): 

     Est.      S.E.    t val.         p
--------- --------- --------- ---------
  0.03080   0.02018   1.52637   0.12705

Slope of fb_fact_check when Consp_Mentality =  0.0000000 (Mean): 

      Est.      S.E.     t val.         p
---------- --------- ---------- ---------
  -0.00191   0.01151   -0.16592   0.86824

Slope of fb_fact_check when Consp_Mentality = -0.6444843 (Low): 

      Est.      S.E.     t val.         p
---------- --------- ---------- ---------
  -0.06120   0.03476   -1.76064   0.07843

██████████████████████████████████████ While Ideology (2nd moderator) = -0.490469778 (Very liberal) █████████████████████████████████████ 

SIMPLE SLOPES ANALYSIS

Slope of fb_fact_check when Consp_Mentality =  0.3555157 (High): 

      Est.      S.E.     t val.         p
---------- --------- ---------- ---------
  -0.00920   0.03993   -0.23034   0.81784

Slope of fb_fact_check when Consp_Mentality =  0.0000000 (Mean): 

     Est.      S.E.    t val.         p
--------- --------- --------- ---------
  0.01548   0.02137   0.72413   0.46906

Slope of fb_fact_check when Consp_Mentality = -0.6444843 (Low): 

     Est.      S.E.    t val.         p
--------- --------- --------- ---------
  0.06020   0.04869   1.23637   0.21644

Warning message:
Johnson-Neyman intervals are not available for factor predictors or moderators. 

... but removing the weights argument from the call to lm() as below...

m_unwtd <- lm(FB_manipulate ~ fb_fact_check
              + Gender + ln_age_years + Edu + RaceEthn + Region + Consp_Mentality + Ideology + PID
              + fb_fact_check*Consp_Mentality*Ideology, data = data.2a)

... and then running sim_slopes() with an unweighted linear model results returns the same simple slopes results as above.

Notably, submitting the weighted and unweighted lm() objects yields different results from interact_plot(). Further, the interaction plot generated by the weighted lm() object matches one generated by a svyglm() object, suggesting that interaction_plot() is incorporating the weights but sim_slopes() is not.

Thank you in advance!