Exact(6)
We also constructed mixed models without the control variables to determine lead coefficients unadjusted for covariates.
Table 1 shows the lead coefficients of the different IQ models.
To examine the effect of simultaneous inclusion of all lead variables, we constructed several mixed models each with only one lead variable and statistically compared those lead coefficients with the lead coefficients of the mixed model with all lead variables.
We then fit separate linear models to the data in each of these ranges and compared the blood lead coefficients for the concurrent blood lead index.
Thereby, in case of nickel all coefficients were higher than that in the control while in case of lead coefficients both below and above that in the control were observed.
After regression diagnostics were examined and homogeneity of the blood lead coefficients across sites was evaluated, the fit of all four measures of blood lead was compared using the magnitude of the model R. The blood lead measure with the largest R (adjusted for the same covariates) was selected a priori as the preferred blood lead index relating blood lead to IQ.
Similar(54)
After the multiple regression models were developed, regression diagnostics were employed to ascertain whether the lead coefficient was affected by collinearity or influential observations (Belsley et al. 1980).
Results were similar to the preferred fixed-effects model, with the random-effects model producing a blood lead coefficient that was 3.7% lower (−2.6 vs. −2.7).
Variables were retained in the final models if they were associated with homocysteine levels or significantly influenced the relation of blood or tibia lead with homocysteine (changed the lead coefficient by more than 10%).
In our analysis, simultaneous inclusion of 6- to 10-year BPb and the remaining BPb reduced the size of the 6- to 10-year lead coefficient without changing its variance, rendering it insignificant.
We also examined the effect of any one site on the overall model by calculating the blood lead coefficient in seven identical models, each omitting one of the seven cohorts (Efron and Tibshirani 1993).
Write better and faster with AI suggestions while staying true to your unique style.
Since I tried Ludwig back in 2017, I have been constantly using it in both editing and translation. Ever since, I suggest it to my translators at ProSciEditing.

Justyna Jupowicz-Kozak
CEO of Professional Science Editing for Scientists @ prosciediting.com