The results were similar to the model with birthweight, and thus are not shown here.
To analyze the time trends for birthweight and length at birth in indigenous and non indigenous newborns, multiple linear regression models were fit, with birthweight (g) and length (cm) as the dependent variables and with socioeconomic and demographic factors (rural area, age and years schooling of the mothers, sex and gestational age of the newborn) as control variables.
In the multivariate model (Table 2), the associations with birthweight (Ptrend<0.001), ethnicity (P<0.001), weight (Ptrend<0.001), height (Ptrend<0.001), exercise (Ptrend<0.01) and number of siblings (Ptrend<0.001) remained significant.
We fit linear models for birthweight and generalized linear models with a Poisson distribution and a log link function for low birthweight and anthropometric measures.
$$\end{document} Birthweight as a continuous variable was analysed using multiple linear regression models; low birthweight and SGA with multiple logistic regression models.
In order to compare birthweight data obtained from health cards with data from maternal recall, all multivariable models were fit for two sets of data: a) children with birthweight data from health cards, and b) children with birthweight data as per maternal recall.
Subsequently, we tested for the interaction effect of maternal smoking, GSTT1 and GSTM1 genes polymorphisms with birthweight by adding all the product terms in the regression models.
In adjusted models, maternal birthweight and postnatal conditional linear/length growth were associated with offspring birthweight.
In the multivariate linear regression model, the 1,5-AG level, either in the second or third trimester, was associated with birthweight independently (p < 0.001) of HbA1c, after adjusting for confounders.
In multilevel regression models of birthweight on the five segregation dimensions, isolation and clustering were significantly associated with birthweight in opposite directions.
