MSFR

author: Roberta De Vito, Alejandra Avalos-Pacheco

Fit the Multi-Study Factor Regression model via the ECM algorithm.

1 Fitting the MSFR model via the ECM Algorithm

The following example illustrates how to fit the MSFR model via the ECM Algorithm, using a data set generating from a simulation scenario.

Getting the data

library(MSFR)
data(Scenario1_MSFR.rda)

Obtaining suitable starting values for model parameters

Then we get inizialization for model parameters, with q=3 common factors and q_s=1 study-specific factors for two studies (S=2).

start_value <- start_msfa(X_s, B_s, p_b, k, j_s, constraint = "block_lower2", method = "adhoc")

Fitting the model via ECM

Now we can proceed for estimating the model parameters via the ECM algorithm

ECM_MSFR <- ecm_msfa(X_s, B_s, start=start_value,  nIt = 10000, trace = FALSE)

The estimated matrix of common loadings can be visualized

Phi <- ECM_MSFR$Phi

The estimated matrix of study-specific loadings can be visualized
```{r get spec, results = FALSE}
Lambda_1 <- ECM_MSFR$Lambda_s[[1]]
Lambda_2 <- ECM_MSFR$Lambda_s[[2]]

The estimated matrix of regression coefficients for covariates effect can be visualized
```{r get cov, results = FALSE}
beta <- ECM_MSFR$$beta