Edge no. | Edge | LEO. NB.OCA | Cor ρ | Path coef | Path SE | Path Z | Model prob | Model df | χ2 stat | RMSEA |
|---|
1 | rs3705921 → Insig1 | | 0.22 | 0.18 | 0.081 | 2.2 | | | | |
2 | rs3670293 → Insig1 | | -0.33 | -0.31 | 0.081 | -3.8 | | | | |
3 | rs3675054 → Dhcr7 | | -0.26 | -0.15 | 0.049 | -3.1 | | | | |
4 |
Insig1 → Dhcr7
| 1.2 | 0.81 | 0.79 | 0.049 | 16.1 | 0.24 | 5 | 6.8 | 0.051 |
5 |
Insig1 → Fdft1
| 1.4 | 0.67 | 0.64 | 0.06 | 10.7 | 0.75 | 5 | 2.7 | 0 |
6 | rs3664397 → Fdft1 | | 0.34 | 0.27 | 0.06 | 4.5 | | | | |
- Using the female mouse liver gene expression data, we report edge scores for the known causal relationships Insig 1 → Dhcr 7 and Insig 1 → Fdft 1 and the other edges depicted in Figure 4. The table represents a condensed summary of the NEO software spreadsheet. The high value of LEO.NB.OCA(Insig 1 → Dhcr 7) = 1.2 suggests that this causal model is 101.2≈ 15.8 times more likely than the next best local model. Similarly, LEO.NB.OCA(Insig 1 → Fdft 1) = 1.4 suggests that the causal model is 25 times more likely than the next best local model. The fourth column reports the marginal Pearson correlation coefficient, while the three path columns (standardized path coefficient, asymptotic standard error, and Z-score for the edge) give details for each individual edge in the SEM models. The last five columns summarize the fits of the two best fitting SEM models shown in Figures 4(b) and (c). The model probability column (Eq. 7) was computed using a central χ2 statistic with the 5 degrees of freedom. The high, non-significant model p-values suggest good fit. The Root Mean Square Error of Approximation (RMSEA) is a standard SEM fit evaluation index that, similar to the χ2 stastic, evaluates the overall fit of the SEM model; a value smaller than 0.05 is desirable.
