. 2019 Mar 12;10:393. doi: 10.3389/fimmu.2019.00393

Table 2.

Potential associations of various fixed effects with immune cell numbers in porcine blood^a.

Variable	Fixed effects: Estimated associations^b with significance levels^c
	Treatment (LD)	Light (off)	Feeding (yes)	Cortisol	Activity	Hematocrit	Sampling
Leukocytes	^↔	^↑^***	^↓^***	^↓^***	^↓^***	^↑^***	^↔
NK cells	^↔	^↓^***	^↓^*	^↑^***	^↑^***	^↑^***	^↓^***
T cells	^↔	^↑^***	^↔	^↓^***	^↓^***	^↑^***	^↔
Monocytes	^↔	^↔	^↓^***	^↓^***	^↓^**	^↑^***	^↓^**
Neutrophils	^↔	^↓^***	^↓^***	^↔	^↑^**	^↑^***	^↓^*
Eosinophils	^↔	^↑^***	^↓^***	^↓^***	^↓^***	^↑^***	^↔
Total Th cells	^↔	^↑^***	^↔	^↓^***	^↓^***	^↑^***	^↓^*
Cytotoxic T cells	^↔	^↑^***	^↔	^↓^**	^↓^***	^↑^**	^↔
γδ T cells	^↔	^↑^***	^↔	^↓^***	^↓^***	^↑^***	^↔
Naive Th cells	^↔	^↑^***	^↔	^↓^***	^↓^***	^↑^***	^↔
Ag-exp. Th cells	^↔	^↔	^↔	^↑^*	^↓^**	^↑^***	^↓^**

Potential associations were assessed by generalized linear mixed model analyses: y_ij = μ + treatment_i + light_ij + concentrate feeding_j + plasma cortisol concentration_ij + relative activity behavior_ij + hematocrit_ij + sampling_j + animal identity_i + ε_ij. Thereby, y_ij represents the cell number/μl blood of the specific immune cell population in porcine blood for an animal i at sampling j, the intercept μ, the factors treatment (LD/SD), light (off/on), and concentrate feeding (yes/no) as well as the covariates plasma cortisol concentration (ng/ml), relative activity behavior (%), hematocrit (%), and sampling (1–26) were set as fixed effects. The factor animal identity (n = 20) was set as random effect with a scaled identity (ID) covariance structure. Sampling was set as repeated effect with a first order autoregressive (AR(1)) residual covariance structure (ε_ij). Animal identity was set as subject. Detailed model results can be found in Table S2.

Estimated association:

^↑

positive,

^↓

negative,

^↔

none.

^***

P < 0.001,

^**

P < 0.01,

P < 0.05.