R - One covariate applies only to one level of a fixed effect in mixed model

Question

I collected a dependent variable (muscle strength) before (PRE) and after (POST) 15 training sessions. Each session was classified in 2 different types of training (Type A and Type B). I would like to investigate if there are differences in my dependent variable in PRE-POST and type of training (Type A and Type B) controlling for two covariates (continue variables) and using subjects as random effect. My problem is the following: while the Covariate A (subjects wellbeing) influences the dependent variable both in PRE and POST, the Covariate B (session training load) only influences the dependent variable at POST. Hence in my dataset the Covariate B at PRE is empty.

Here the description of my variables: Dependent variable: DV (subjects muscle strength) Independent variable: Type of practice (2 levels: Type A and Type B) Independent variable: Time (2 levels: PRE and POST) Covariate A: continuous variable (subjects wellbeing) Covariate B: continuous variable (session training load) ID: Subjects

I am attempting to fit a mixed effect model using R and lme4, but I am new to the use of mixed models. My model is the following:

Model = lmer (DV~ Time * Type + (Covariate A)+( Covariate B)+(1|subject1) , data=dataset)

dataset looks like this:

ID  Session Time    Type    Covariate A Covariate B DV
Sub1    n1  PRE      A          10                  120
sub1    n1  POST     A          10          5       115
sub1    n2  PRE      B          14                  122
sub1    n2  POST     B          14          6       124
sub1    n3  PRE      A          12                  123
sub1    n3  POST     A          12          7       119
sub1    n4  PRE      B          9                   118
sub1    n4  POST     B          9           8       120
…       …    …       …          …           …        …

However, when I run the model with this database, I receive the following error: Error in contrasts<-(*tmp*, value = contr.funs[1 + isOF[nn]]) : contrasts can be applied only to factors with 2 or more levels.`

Differently, If I run the analysis with the following data set, the model works. However, I know that the Covariate B does not influence the DV in the PRE. Thus it is illogical to apply a value of covariate B to an observation at PRE.

ID  Session Time    Type    Covariate A Covariate B DV
Sub1    n1  PRE      A          10          5       120
sub1    n1  POST     A          10          5       115
sub1    n2  PRE      B          14          6       122
sub1    n2  POST     B          14          6       124
sub1    n3  PRE      A          12          7       123
sub1    n3  POST     A          12          7       119
sub1    n4  PRE      B          9           8       118
sub1    n4  POST     B          9           8       120
…       …    …       …          …           …        …

How Can organize the database considering this aspect?

---

Following your suggestions, here you can find an example that you can run: drive.google.com/open?id=1h1F_ITZmmCN5sNKtMsETM2yakx3Kjmn0

Model 1: the right model where the CovariateB (session training load) only influences the dependent variable at POST.

Model 2: the wrong model where the CovariateB_2 (session training load) is reported both in PRE and POST.

R script:

library(lme4)
library(lmerTest)
library(emmeans)
library(readxl)

dataset = read_excel("$$:/$$$/$$/$$.xlsx", sheet = NULL, range = NULL, col_names = TRUE)

dataset$Time <- as.factor(dataset$Time)
dataset$Type <- as.factor(dataset$Type)
dataset$Subject <- as.factor(dataset$Subject)



Model_1 = lmer(((DV)) ~ Time*Type+(CovariateA)+(CovariateB)+(1|Subject), data=dataset, REML=FALSE);
Model_2 = lmer(((DV)) ~ Time*Type+(CovariateA)+(CovariateB_2)+(1|Subject), data=dataset, REML=FALSE);