Excerpts

At office level, only one random coefficient for the constant turned out to be weakly significant (P ≤ 0.10; for NCSE), indicating that there was a small systematic difference for this outcome measure between offices.

In these tests age, sex, and co-morbidity will be included as covariates, in addition to those other potentially confounding variables that will show an at least marginally significant (p < .10) difference between groups.

In Model I, the mediation effect was not significant (B = 0.004, t = 0.888), while in Model II, the mediation effect was marginally significant (B = 0.012, t = 1.596, p < 0.10).

The variables that were marginally significant with P < 0.10 in the univariate analysis were eligible for entry into the multivariable logistic regression.

However, in Model III, the mediation effect approached significance (B = 0.043, t = 1.330, p < 0.10), suggesting that in low-density, large campuses, convenience may play a more substantial role in mediating the relationship between depression and perceived restorative effects.

Significance was set at p < 0.05, and borderline significance at 0.05 < p < 0.10.

The combined effect of the two variables, just marginally non-significant ( p < .10), is smaller in magnitude than the independent effect of deprivation, reflecting the statistical averaging of the two effects, contrary to Hypothesis 1a.

Because the sample size is small, associations with borderline significance (0.5 < p < 0.10) are also reported.

Multivariate analysis of significant or nearly significant effects (p < 0.1) of lifestyle variables concerned in this study (maternal age over 35 or under 17, being single mother, primiparity, smoking during pregnancy, history of infertility, previous miscarriages, previous induced abortions, short time or long time since previous pregnancy or using IUD before this pregnancy) was based on multiple logistic regression analysis (BMDP Statistical Software Inc., Los Angeles, CA).

Heating mode and recent painting had effects of borderline significance (0.05 < p < 0.10) on personal exposure to toluene and xylene (Figure 6 ).

Adults with hypertension scored on average 2.32 points lower (95% CI: -4.84–0.21) on physical health QoL compared to the healthy group and this difference had borderline significance ( p < 0.10).

Significant ( p < .05) or marginally significant ( p < .10) differences existed on parental BMI for parents’ and children’s steps/day.

The results show a statistically significant decline over time in the mean SBP both among women aged 55 years and older of 0.8 mmHg per year ( p < 0.01) and a marginally significant decline of 0.25 mmHg per year in the younger age group 35–54 ( p < 0.1).

At one standard deviation below the mean of perceived atmosphere, the change between T1 and T2 is negative for the psychological and the physical domains (estimate = -.106, p = .009 and estimate = -.074, p = .10; marginal significance), but not for the social and the environmental domains (estimate = -.141, p = .11 and estimate -.009, p = .867, respectively) (Fig. 2 ).

For both multivariate logistic regressions, we retained significant or near-significant univariate predictors ( p < .10), and we forced the following variables into the model: age, food insecurity, and number of children.

The size of the full yingchou network was still marginally significant ( p < 0.10) in Model 3, and participants’ age was significant ( p < 0.05), with older participants less likely to report commercial sex in the past year.

For all analyses, the significance was set at p < 0.05, and a trend towards significance was noted when p < 0.1.

Statistical significance was set at the p = 0.05 level and borderline significance was set at p ≤ 0.10 for all tests.

Variables that were significant or approaching significance with p ≤0.10 in their respective univariate models were considered as covariates in the multivariable models.

The influence of the level of breaks at work and attitudes towards interrupting sitting at work was borderline significant (P < .10; see Table 3 ).

All covariates were tested in the model but only those that achieved or approached significance (p < 0.10) were retained.

Age showed a nonsignificant trend towards lower stress at higher age (B = − 0.04, p = 0.106).

Despite approaching significance, bivariate analyses did not reveal a significant difference in criminogenic risk levels ( p = .11, Cramer’s V = 0.24).

Although Ace and Ant became marginally significant ( P = 0.111 and 0.096 , respectively), likely owing to reduced statistical power from the smaller number of incident cases at the higher threshold, their effect directions and magnitudes (ORs) remained stable.

In contrast, the marginally non-significant p-value for the mental health index ( p = 0.12) indicated that OLS and IV estimates did not differ substantially for this outcome, implying weaker endogeneity concerns.