# Extract the community matrix (B) Extract the community matrix (B) estimated by a MARSS model. The matrix is necessary to calculate the Jacobian metrics. ```r extractB(marss_res, states_names = NULL) ``` ## Arguments - `marss_res`: MARSS object returned by `MARSS` - `states_names`: a string vector containing the names of species/groups for which interactions were estimated. ## Returns A named square matrix (B) with one row (and column) per species/group listed in `states_names` (which should also have been used in `MARSS`). Row and column names are included (named after `states_names`). The values are interaction strengths between the species/groups estimated by MARSS. ## Examples ```r library(MARSS) # smaller dataset for example: # 3 functional groups and two insecticide concentrations besides control data_df <- subset( aquacomm_fgps, treat %in% c(0.0, 0.9, 6) & time >= 1 & time <= 28, select = c(time, treat, repli, herb, carn, detr) ) # estimate z-score transformation and replace zeros with NA data_df[, c("herb", "carn", "detr")] <- lapply(data_df[, c("herb", "carn", "detr")], MARSS::zscore) data_df[, c("herb", "carn", "detr")] <- lapply(data_df[, c("herb", "carn", "detr")], function(x) replace(x, x == 0, NA)) # reshape data from wide to long format data_z_ldf <- reshape( data_df, varying = list(c("herb", "carn", "detr")), v.names = "abund_z", timevar = "fgp", times = c("herb", "carn", "detr"), direction = "long", idvar = c("time", "treat", "repli") ) data_z_ldf <- data_z_ldf[order(data_z_ldf$time, data_z_ldf$treat, data_z_ldf$fgp),] # summarize mean and sd data_z_summldf <- aggregate(abund_z ~ time + treat + fgp, data_z_ldf, function(x) c(mean = mean(x, na.rm = TRUE), sd = sd(x, na.rm = TRUE))) data_z_summldf <- do.call(data.frame, data_z_summldf) names(data_z_summldf)[4:5] <- c("abundz_mu", "abundz_sd") # split dataframe per functional groups # into list to apply the MARSS model more easily split_data_z <- split(data_z_summldf[, c("time", "fgp", "abundz_mu")], data_z_summldf$treat) reshape_to_wide <- function(df) { df_wide <- reshape(df, idvar = "fgp", timevar = "time", direction = "wide") rownames(df_wide) <- df_wide$fgp df_wide <- df_wide[, -1] # Remove the 'fgp' column as.matrix(df_wide) } data_z_summls <- lapply(split_data_z, reshape_to_wide) # fit MARSS models data.marssls <- list( MARSS( data_z_summls[[1]], model = list( B = "unconstrained", U = "zero", A = "zero", Z = "identity", Q = "diagonal and equal", R = matrix(0, 3, 3), tinitx = 1 ), method = "BFGS" ), MARSS( data_z_summls[[2]], model = list( B = "unconstrained", U = "zero", A = "zero", Z = "identity", Q = "diagonal and equal", R = matrix(0, 3, 3), tinitx = 1 ), method = "BFGS" ), MARSS( data_z_summls[[3]], model = list( B = "unconstrained", U = "zero", A = "zero", Z = "identity", Q = "diagonal and equal", R = matrix(0, 3, 3), tinitx = 1 ), method = "BFGS" ) ) # identify experiments names(data.marssls) <- paste0("Conc. = ", c("0", "0.9", "44"), " micro g/L") # extract community matrices (B) data.Bls <- data.marssls |> lapply(extractB, states_names = c("Herbivores", "Carnivores", "Detrivores")) print(data.Bls) ``` ## References Holmes, E. E., Ward, E. J., Scheuerell, M. D., & Wills, K. (2024). MARSS: Multivariate Autoregressive State-Space Modeling (Version 3.11.9). tools:::Rd_expr_doi("10.32614/CRAN.package.MARSS") Holmes, E. E., Scheuerell, M. D., & Ward, E. J. (2024). Analysis of multivariate time-series using the MARSS package. Version 3.11.9. tools:::Rd_expr_doi("10.5281/zenodo.5781847") Holmes, E. E., Ward, E. J., & Wills, K. (2012). MARSS: Multivariate autoregressive state-space models for analyzing time-series data. The R Journal, 4(1), 30. tools:::Rd_expr_doi("10.32614/RJ-2012-002")

extractB function