home.bias = function(own.home.assets, own.all.assets, others.home.assets, others.all.assets){
  1 - (1-own.home.assets/own.all.assets)/(1-others.home.assets/others.all.assets)
}

home.bias.subtract = function(own.home.assets, own.all.assets, others.home.assets, others.all.assets){
  own.home.assets/own.all.assets - others.home.assets/others.all.assets
}


home.bias.permutation = function(type = "state", M = 1000, skip.biases = FALSE, category = "assets"){

  if(type == "state"){
    prefix = ""
    id.var = "state"
    first.colnum.of.patterns = 10
  }else if(type == "committee"){
    prefix = "scb."
    id.var = "sic.code.bunch"
    first.colnum.of.patterns = 11
  }else if (type == "contributions"){
    first.colnum.of.patterns = 11
    prefix = "pia."
    id.var = "pac.bunch"
  }
  
    # get data and get rid of people with no trades
    a.all = read.csv(paste(type, "_full_matrix_", category, ".csv", sep = ""), header = T) # this is in summary_stats 
    a = a.all[which(a.all$total.weight > 0), ,]

    # grab the names of the patterns as they appear in the columnames of a  
    pattern.names = colnames(a)[first.colnum.of.patterns:ncol(a)]
    ####

    # we calculate the home bias by pattern (state, pac bunch, committee bunch) and member once -- what the home bias would be for this portfolio/member for each state/etc considered.
    biases = matrix(NA, nrow = nrow(a), ncol = length(pattern.names)) # the holder: every guy and pattern for which we counted assets.
    colnames(biases) = pattern.names # eg "NY", scb.1
    rownames(biases) = a$name  # eg. Barney Frank
    # biases.subtract = biases

    cat("Calculating the biases under the various scenarios. For", nrow(biases), "members.\n")

    # get holdings and biases by pattern  
    for (i in 1:nrow(biases)){ # for each member
      if(i %% 10 == 0){cat(i, "")}
      for (j in 1:ncol(biases)){
        if(skip.biases == TRUE){next}                                # for each column -- one per bundle of connections for which we have calculated assets 
        own.colname = pattern.names[j] # paste(prefix, a[[id.var]][j], sep = "") # the colname of the connected pattern for this iteration -- eg. "NY", "scb.12", "pia.12"
        # cat("Own.colname is", own.colname, "\n")
        own.in = a[i, own.colname]  # the value in the input for this guy on that column
        all.in = sum(a[, own.colname])  # the value for everyone on that column 
        biases[i,j] = home.bias(own.in, a[i,"total.weight"], all.in - own.in,
                sum(as.numeric(a$total.weight[-i])))
      #   biases.subtract[i,j] = home.bias.subtract(own.in, a[i,"total.weight"], all.in - own.in,
              #   sum(as.numeric(a$total.weight[-i])))  
      }
    }
    # this takes a little while.

    # now we permute the patterns
    n = nrow(biases) # number of guys
    out = matrix(NA, ncol = n, nrow = M)  # storage matrix for permutations
    colnames(out) = a$name
    out.subtract = out
  
    permutation = a[[id.var]] # the column with the assignments, which we permute
    cat("Going through", M, "permutations.\n")

    for (m in 1:M){  # through permutations 
      if (m %% 10 == 0){cat(m, " ")}
      if (m > 1){permutation = sample(a[[id.var]])}  # we do a thought experiment of shuffling the patterns
      for (i in 1:n){ # cycling through guys
        
        this.column = paste(prefix, permutation[i], sep = "") # the counterfactual bundle drawn by the i'th guy in this permutation
        out[m,i] = biases[i,this.column]  # take the bias of the i'th guy given the counterfactual bundle from this permutation. 
       #  out.subtract[m,i] = biases.subtract[i,this.column]  # take the bias of the i'th guy given the counterfactual bundle from this permutation. 
      }
    }
    print("Done, motherfucker.")

    bias = data.frame(id = a$id, name = a$name, bias = t(t(out[1,]))) 
  #   bias.subtract = data.frame(id = a$id, name = a$name, bias = t(t(out.subtract[1,]))) 
      
    list(sims = out, type = type, M = M, bias = bias) # , bias.subtract = bias.subtract) sims.subtract = bias.subtract,

}

# just two columns: the member's in-group avg and everyone else's average in that group

home.bias.simple = function(type = "state", M = 1000, skip.biases = FALSE, category = "assets"){

  if(type == "state"){
    prefix = ""
    id.var = "state"
    first.colnum.of.patterns = 10
  }else if(type == "committee"){
    prefix = "scb."
    id.var = "sic.code.bunch"
    first.colnum.of.patterns = 11
  }else if (type == "contributions"){
    first.colnum.of.patterns = 11
    prefix = "pia."
    id.var = "pac.bunch"
  }
  
    # get data and get rid of people with no trades
    a.all = read.csv(paste(type, "_full_matrix_", category, ".csv", sep = ""), header = T) # contributions
    a = a.all[which(a.all$total.weight > 0), ,]

    # grab the names of the patterns as they appear in the columnames of a  
    # pattern.names = as.character(a[[id.var]])
# colnames(a)[first.colnum.of.patterns:ncol(a)]
 #    if(type == "state"){pattern.names = as.character(a$state)}
    ####

    # we calculate the home bias by pattern (state, pac bunch, committee bunch) and member once -- what the home bias would be for this portfolio/member for each state/etc considered.
    biases = matrix(NA, nrow = nrow(a), ncol = 4) # the holder: for every guy we get the sum in and out, for self and others
    colnames(biases) = c("in.self", "total.self", "in.others", "total.others") # pattern.names # eg "NY", scb.1
    rownames(biases) = a$name  # eg. Barney Frank

    cat("Calculating the in and out sums, for self and others, for", nrow(biases), "members.\n")

    # get holdings and biases by pattern  
    for (i in 1:nrow(biases)){ # for each member
      if(i %% 10 == 0){cat(i, "")}
      own.colname = paste(prefix, a[[id.var]][i], sep = "") #  pattern.names[i] # the colname of the connected pattern for this iteration -- eg. "NY", "scb.12", "pia.12"
      own.in = a[i, own.colname]  # the value in the input for this guy on that column
      all.in = sum(a[, own.colname])  # the value for everyone on that column 
      own.total = a[i, "total.weight"]
      all.total = sum(as.numeric(a$total.weight[-i]))
      biases[i,] = c(own.in, own.total, all.in, all.total)
      }
  biases
}



loc.bias = function(affiliation.index, own.holdings.index, df = a, house.inside = "state"){
  # get own associations
  ai = affiliation.index
  # aff = df[affiliation.index,c(5, 7, 8)]
  chamber = as.character(df[ai, "chamber"])
  state = as.character(df[ai, "state"])
  district = as.character(df[ai, "district"])
  # identify the columns for which the holdings are inside, given this affiliation index
  if(house.inside == "state"){  # can eventually do district bias
    own.inside.columns = grep(state, names(df))  }
  #print("got to here 1")
  # print(own.inside.columns)
  own.inside.holdings = sum(df[own.holdings.index, own.inside.columns])
  own.total.holdings = sum(df[own.holdings.index, 9:ncol(df)])
  #print("got to here 2")
  others.inside.holdings = sum(df[-own.holdings.index, own.inside.columns])
  others.total.holdings = sum(df[-own.holdings.index, 9:ncol(df)])
  #print("got to here")
  home.bias(own.inside.holdings, own.total.holdings, others.inside.holdings, others.total.holdings)
}