To overcome the sparse data problem in regression, we may fit a conditional logistic regression model, in which the intercept, which is a nuisance parameter in effect estimation, is not estimated.
More formally, given genotype data from "duos" consisting of offspring (affected or unaffected) together with their mothers, one may fit models (via logistic regression, for example) that incorporate effects of offspring genotype, maternal genotype, maternal-fetal interactions, and imprinting [ Chen et al., 2009; Li et al., 2009; Shi et al., 2008; Weinberg and Umbach, 2005].
The lack of fit is not significant (p > 0.05), which means the regression equation may fit the actual situation.
The slightly different curvature of the means may reflect that the fitted regression line would be different for the short time span of the present data compared the longer span of the other two studies.
It is possible that the fitted regression model may be biased towards strong binding sites.
Although some observations on the scatter plots lie outside the CIs, these may have minimal impact on the fitted regression if the sample sizes are relatively small, as these are weighted regressions.
In fact, a fitted regression line shows the level of support falling slightly during the last three decades.
Fig. 4 By-trial fitted regression lines.
However, the fitted regression line had an R = 0.977.
Assuming a linear relationship between porosity and AI, a straight line may be fitted by regression in Fig. 3.
