docs

PharmCat · Mar 1, 2021 · 714bd6c · 714bd6c
1 parent 0b2727e
commit 714bd6c
Show file tree

Hide file tree

Showing 7 changed files with 20 additions and 12 deletions.
diff --git a/docs/src/api.md b/docs/src/api.md
@@ -16,6 +16,11 @@
 
 [`fit!`](@ref)
 
+### Metida.CovarianceType
+```@docs
+Metida.CovarianceType
+```
+
 ### Metida.ScaledIdentity
 ```@docs
 Metida.ScaledIdentity

diff --git a/docs/src/examples.md b/docs/src/examples.md
@@ -43,6 +43,7 @@ MixedModels result:
 ```@example lmmexample
 fm = @formula(response ~ 1 + factor*time + (1 + time|subject&factor))
 mm = fit(MixedModel, fm, rds, REML=true)
+println(mm) #hide
 ```
 
 ### Example 2 - Two random factors (Penicillin data)
@@ -66,6 +67,7 @@ MixedModels:
 
 fm2 = @formula(diameter ~ 1 + (1|plate) + (1|sample))
 mm = fit(MixedModel, fm2, df, REML=true)
+println(mm) #hide
 ```
 
 ### Example 3 - Repeated ARMA/AR/ARH

diff --git a/docs/src/faq.md b/docs/src/faq.md
@@ -8,6 +8,10 @@ Metida.jl work with small and medium datasets. Model fitting is based on varianc
 
 Blocks depend on subjects. If you have only one random effect block is equivalent to subject. If you have more than one random effect blocks will be made as non-crossing combination for all subject variables.     
 
-* Q3: model not converging!
+* Q3: Why model does not converge?
 
 Optimization of REML function can depend on many factors. In some cases covariance parameters can be correlated (ill-conditioned/singular covariance matrix). So hypersurface in the maximum area can be very flat, that why the result can be different for different starting values (or for different software even REML is near equal). Also, some models can not be fitted for specific data at all. If the model not fitted try to check how meaningful and reasonable is the model or try to guess more robust initial conditions.
+
+* Q2: Is model fitting is correct?
+
+Use 'lmm.log' to see warnings and errors. If you have warnings and error maybe this model is overspecified. Try to change variance structure or guess  more robust initial conditions. 
diff --git a/docs/src/instanduse.md b/docs/src/instanduse.md
@@ -94,13 +94,15 @@ Metida.fit!
 #### Step 1: make CustomCovarianceStruct
 
 ```@docs
-Metida.CustomCovarianceStruct
+Metida.CovmatMethod
 ```
 
-#### Step 2: make CustomCovarianceType
+See: [`Metida.CovmatMethod`](@ref)
 
-```@docs
-Metida.CustomCovarianceType
+#### Step 2: make custom CovarianceType
+
+```
+  CovarianceType(cm::AbstractCovmatMethod)
 ```
 
 #### Step 3 Fit your model

diff --git a/docs/src/validation.md b/docs/src/validation.md
@@ -238,11 +238,6 @@ fit!(lmm)
 | 7 | plate   σ² | 0.716908 |  |  |
 | 7 | sample   σ² | 3.73092 |  |  |
 | 7 | Residual   σ²| 0.302415 |  |  |
-|    |  |  |  |  |
-|    |  |  |  |  |
-|    |  |  |  |  |
-|    |  |  |  |  |
-|    |  |  |  |  |
 
 ## Section 3: Validation with bioequivalence datasets with Metida & SPSS
 

diff --git a/src/Metida.jl b/src/Metida.jl
@@ -26,7 +26,7 @@ TOEP, Toeplitz,
 TOEPP, ToeplitzParameterized,
 TOEPH, HeterogeneousToeplitz,
 TOEPHP, HeterogeneousToeplitzParameterized,
-CovmatMethod,
+CovarianceType, CovmatMethod,
 fit!, LMM, VarEffect, theta, logreml, m2logreml, thetalength, dof_satter, dof_contain, rankx, caic,
 gmatrix, rmatrix, vmatrix!,
 AbstractCovarianceType, AbstractCovmatMethod, MetidaModel

diff --git a/src/varstruct.jl b/src/varstruct.jl
@@ -282,7 +282,7 @@ end
 """
     CovarianceType(cm::AbstractCovmatMethod)
 
-Make custom covariance type with CustomCovarianceStruct.
+Make custom covariance type with CovmatMethod.
 """
 CovarianceType(cm::AbstractCovmatMethod) = CovarianceType(:FUNC, cm)