association of epigenetic age with categorical variable

Question

I am looking for association between road buffer zone and epigenetic age acceleration? not sure if this is a correct approach?

 > model1 <- lm (api.age ~ buffer.zone + Smoking_Status + sex  + Age, data = ph1)
> 
> summary( moddel1)

Call:
lm(formula = api.age ~ buffer.zone + Smoking_Status + 
    sex + Age, data = ph1)

Residuals:
     Min       1Q   Median       3Q      Max 
-16.4799  -2.8949  -0.4211   2.5132  24.1447 

Coefficients:
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)      0.382568   0.982393   0.389  0.69703    
buffer.zone300m -0.269675   0.421202  -0.640  0.52213    
Smoking_Status   0.786199   0.198754   3.956 8.06e-05 ***
sex          -0.784025   0.263335  -2.977  0.00296 ** 
Age           -0.003015   0.013610  -0.222  0.82470    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 4.632 on 1269 degrees of freedom
Multiple R-squared:  0.02192,   Adjusted R-squared:  0.01883 
F-statistic: 7.108 on 4 and 1269 DF,  p-value: 1.167e-05

> # Coefficient estimates
> coefficients <- coef(model1)
> coefficients 
    (Intercept) buffer.zone300m  Smoking_Status          sex           Age 
    0.382568307    -0.269674876     0.786199085    -0.784025407    -0.003015226 
> # Confidence intervals
> conf_intervals <- confint(model1)
> conf_intervals
                      2.5 %      97.5 %
(Intercept)     -1.54472463  2.30986125
buffer.zone300m -1.09600370  0.55665395
Smoking_Status   0.39627598  1.17612219
sex          -1.30064500 -0.26740582
Age           -0.02971586  0.02368541
> # Combine coefficients and confidence intervals
> combined_output <- cbind(coefficients, conf_intervals)
> combined_output
                coefficients       2.5 %      97.5 %
(Intercept)      0.382568307 -1.54472463  2.30986125
buffer.zone300m -0.269674876 -1.09600370  0.55665395
Smoking_Status   0.786199085  0.39627598  1.17612219
sex          -0.784025407 -1.30064500 -0.26740582
Age           -0.003015226 -0.02971586  0.02368541

Many thanks,

Answer 1

I would personally first check what is the distribution of the response variable.

It looks like there is no simple effect of the size of the buffer zone on the epigenetic age acceleration. 1000 m is the intercept.

I would personally wonder whether there is an interaction between your explanatory variables : model1 <- lm (PCGrimAge ~ buffer.zone * smoking_status * sex + Age, data = ph) Then I would check whether model assumptions are met with the plot(simulateResiduals(model1)) of the DHARMa package. If yes, I would use the dredge function of the MuMIn package to see what variables/interactions receive support in the models within a delta AIC of 2.