3
$\begingroup$

I am using a mixed model to predict the effect of certain environmental exposures on brain region measures. So temp[, j] gives the brain region our regressor and temp[, var] gives the environmental variable of the predictor. Height, weight, intracranial volume, age are controlled for and random effects for family relation and scanner site are controlled for. The terms that matter are temp[,var] and its interactions with sex and race_ethnicity.

fml<-lmer(temp[,j]~ interview_age+rel_relationship+ehi_y_ss_scoreb+
          scale(smri_vol_scs_intracranialv)+
          (1|mri_info_deviceserialnumber)+ 
          (1|mri_info_deviceserialnumber:rel_family_id)+
           anthroheightcalc + anthroweightcalc + sex + race_ethnicity +
           temp[,var] + temp[,var]:sex + temp[,var]:race_ethnicity,
           data = temp)

We are interested in how race as a social construct and gender may influence them so they have been modelled. sex has two levels race_ethnicity in our case has 3 levels. The coefficients for the fixed effects are:

> coef(summary(fml))
                                      Estimate Std. Error        df    t value  Pr(>|t|)
(Intercept)                       4207.6374706 80.6146186 8102.8607 52.1944722 0.0000000
interview_age                       -0.6941963  0.5300313 8904.7643 -1.3097271 0.1903220
rel_relationship                    -2.8646617  4.6742776  219.7843 -0.6128566 0.5406048
ehi_y_ss_scoreb                     -1.7937162  5.0437752 9048.6169 -0.3556297 0.7221262
scale(smri_vol_scs_intracranialv)  280.7948583  3.5021661 9074.3988 80.1774825 0.0000000
anthroheightcalc                    -0.5141972  1.6238005 8889.1745 -0.3166628 0.7515069
anthroweightcalc                     0.1612428  0.2037307 8738.9575  0.7914507 0.4287025
sexM                                 4.4027100 12.5976755 9078.9926  0.3494859 0.7267327
race_ethnicity2                      5.0497778 22.2093470 5893.2108  0.2273717 0.8201426
race_ethnicity3                     14.9754508 16.2265245 2172.1264  0.9228995 0.3561621
temp[, var]                          0.1745047  0.2581049 1963.3899  0.6760999 0.4990568
sexM:temp[, var]                    -0.3622916  0.2689064 9076.3253 -1.3472776 0.1779245
race_ethnicity2:temp[, var]         -0.1747920  0.3647274 7669.2146 -0.4792401 0.6317815
race_ethnicity3:temp[, var]         -0.2288939  0.3619681 4730.3581 -0.6323591 0.5271828

What does the intercept represent in this case? Is it the the mean of temp[, var] at sex F and race_ethnicity 1? Then the coefficients are deviations from this grand mean?

I think the answers in the posts here and here are relevant but given the more complex design of my model I think this warrants a separate answer.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

What does the intercept represent in this case? Is it the the mean of temp[, var] at sex F and race_ethnicity 1? Then the coefficients are deviations from this grand mean?

It is the expected value of the response at the reference level of all the categorical variables, and at zero for all numeric variables.

Improve this answer
$\endgroup$
4
  • $\begingroup$ So it it possible for me to extract a coeeficient for sexF:temp[, var] or for race_ethncity:temp[,var]? $\endgroup$
    – Angus Campbell
    Commented Jul 7, 2021 at 21:56
  • $\begingroup$ It's not clear what you mean by that. But it would be a seperate question anyway. However, you might want to look into lsmeans/emmeans which may be what you want. $\endgroup$
    – Robert Long
    Commented Jul 7, 2021 at 22:10
  • $\begingroup$ Does this answer your question ? If so please consider marking it as the accepted answer. If not, please let us know why. Also, if you haven't already, please consider upvoting it. $\endgroup$
    – Robert Long
    Commented Jul 24, 2021 at 12:05
  • 1
    $\begingroup$ I upvoted it sorry I kind of forgot about this question. Thanks for the gentle reminder. I have accepted it now. $\endgroup$
    – Angus Campbell
    Commented Jul 24, 2021 at 17:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.