Read in the data and create separate slope variables and obsid variable.
library(tidyr)
library(dplyr)
library(nlme)
kashy_ppp <- read.csv(file.choose(), header=TRUE)
kashy_ppp <- kashy_ppp %>%
mutate(slope_m = man*(time), slope_w = woman*(time), obsid = Day+14*(dyadid-1))
APIM_long <- lme(satisf_A ~ genderE + conflict_A + conflict_P
+ genderE*conflict_A + genderE*conflict_P,
data = kashy_ppp,
random = ~ man + woman + conflict_A + conflict_P - 1|dyadid,
correlation = corCompSymm(form = ~1|dyadid/obsid),
weights = varIdent(form = ~1|genderS),
na.action = na.omit)
summary(APIM_long)
## Linear mixed-effects model fit by REML
## Data: kashy_ppp
## AIC BIC logLik
## 5134.899 5247.891 -2548.449
##
## Random effects:
## Formula: ~man + woman + conflict_A + conflict_P - 1 | dyadid
## Structure: General positive-definite, Log-Cholesky parametrization
## StdDev Corr
## man 0.60782469 man woman cnfl_A
## woman 0.48142010 0.756
## conflict_A 0.11813139 -0.086 -0.147
## conflict_P 0.08352535 -0.245 -0.419 0.529
## Residual 0.52654298
##
## Correlation Structure: Compound symmetry
## Formula: ~1 | dyadid/obsid
## Parameter estimate(s):
## Rho
## 0.2570029
## Variance function:
## Structure: Different standard deviations per stratum
## Formula: ~1 | genderS
## Parameter estimates:
## M F
## 1.000000 1.041919
## Fixed effects: satisf_A ~ genderE + conflict_A + conflict_P + genderE * conflict_A + genderE * conflict_P
## Value Std.Error DF t-value p-value
## (Intercept) 6.793863 0.05703066 2725 119.12649 0.0000
## genderE -0.105341 0.02716656 2725 -3.87759 0.0001
## conflict_A -0.162913 0.01535975 2725 -10.60651 0.0000
## conflict_P -0.067184 0.01275868 2725 -5.26576 0.0000
## genderE:conflict_A 0.023139 0.01104504 2725 2.09494 0.0363
## genderE:conflict_P -0.005632 0.01091235 2725 -0.51610 0.6058
## Correlation:
## (Intr) gendrE cnfl_A cnfl_P gnE:_A
## genderE 0.205
## conflict_A -0.263 0.021
## conflict_P -0.406 0.085 0.205
## genderE:conflict_A -0.013 -0.235 0.040 0.032
## genderE:conflict_P 0.023 -0.232 -0.014 -0.084 -0.700
##
## Standardized Within-Group Residuals:
## Min Q1 Med Q3 Max
## -8.21124048 -0.32191033 0.08574289 0.42395256 4.04125601
##
## Number of Observations: 2833
## Number of Groups: 103
Create lagged variables.
kashy_ppp <- kashy_ppp %>%
group_by(dyadid, person) %>%
mutate(conflict_A_lag = lag(conflict_A),
conflict_P_lag = lag(conflict_P))
Use the lagged actor and partner variables. Note: the random effects of the lagged vairables could not be estimated with default iteration criteria.
stability_influence <- lme(satisf_A ~ genderE + conflict_A_lag + conflict_P_lag
+ genderE*conflict_A_lag + genderE*conflict_P_lag,
data = kashy_ppp,
random = ~ man + woman + slope_m + slope_w - 1|dyadid,
correlation = corCompSymm(form = ~1|dyadid/obsid),
weights = varIdent(form = ~1|genderS),
na.action = na.omit)
summary(stability_influence)
## Linear mixed-effects model fit by REML
## Data: kashy_ppp
## AIC BIC logLik
## 5227.887 5339.442 -2594.944
##
## Random effects:
## Formula: ~man + woman + slope_m + slope_w - 1 | dyadid
## Structure: General positive-definite, Log-Cholesky parametrization
## StdDev Corr
## man 0.68670843 man woman slop_m
## woman 0.54831053 0.799
## slope_m 0.06050170 -0.096 0.077
## slope_w 0.05550318 -0.025 0.112 0.370
## Residual 0.56848033
##
## Correlation Structure: Compound symmetry
## Formula: ~1 | dyadid/obsid
## Parameter estimate(s):
## Rho
## 0.4424515
## Variance function:
## Structure: Different standard deviations per stratum
## Formula: ~1 | genderS
## Parameter estimates:
## M F
## 1.00000 1.10986
## Fixed effects: satisf_A ~ genderE + conflict_A_lag + conflict_P_lag + genderE * conflict_A_lag + genderE * conflict_P_lag
## Value Std.Error DF t-value p-value
## (Intercept) 6.389147 0.06557319 2519 97.43535 0.0000
## genderE -0.062342 0.02754897 2519 -2.26294 0.0237
## conflict_A_lag -0.010262 0.00993596 2519 -1.03283 0.3018
## conflict_P_lag -0.019723 0.00999770 2519 -1.97280 0.0486
## genderE:conflict_A_lag -0.004603 0.01249644 2519 -0.36831 0.7127
## genderE:conflict_P_lag 0.009993 0.01254681 2519 0.79648 0.4258
## Correlation:
## (Intr) gendrE cnf_A_ cnf_P_ gE:_A_
## genderE 0.184
## conflict_A_lag -0.291 0.029
## conflict_P_lag -0.291 0.076 -0.049
## genderE:conflict_A_lag -0.008 -0.206 0.015 0.085
## genderE:conflict_P_lag 0.038 -0.210 -0.007 -0.183 -0.771
##
## Standardized Within-Group Residuals:
## Min Q1 Med Q3 Max
## -6.5246302 -0.2164129 0.1083552 0.3976447 3.1681865
##
## Number of Observations: 2627
## Number of Groups: 103