Among the studies that have dealt with ozone as a third variable in the heat mortality association, the motivations and explanations vary.
This association could not obviously be attributed to an indirect influence through a third variable, since letter length was not correlated with any of the other predictor variables.
Moreover, an association with clinical variables and cytogenetic aberrations has been proposed8,9,10.
Indeed, even if carry-over effects existed, there is no reason to assume that these should per se affect the reported associations with third variables.
In a first analysis of the proximate variables investigated, ANC attendance, delivery place, and distance to the closest health facility showed an association with neonatal mortality.
First, a consistent challenge in genetic association studies are that of third variable confounds, or unmeasured variables that are causally responsible for the observed finding but are associated with the measured variables thus generating a spurious association.
In a second step, fixed level 1 and level 2 variables were added and tested for their association with the dependent variable, the PACIC score.
A second multivariable logistic regression model (full model) was then conducted by also including attitudinal and behavioral variables as candidate variables that showed marginally statistically significant association with the outcome variable (p < 0.15) in unadjusted analyses (see Table 4).
From two models, variables which had significant association with dependent variable (P < 0.05) were entered to the final third regression model.
Independent variables were first evaluated for unconditional associations with the dependent variable using a chi-square test for categorical data and t test for continuous data.
We analyzed their association with modifiable perioperative variables.
