In USEPA/stateio: US State Input-Output (stateio) R modeling software
This document presents validation results of r year summary-level state IO model.
knitr::opts_chunk$set(echo = FALSE,
message = FALSE,
results = "asis")
logging::basicConfig(level = 40)
library(devtools)
load_all()
Prepare data
0. Load data
source('DefineMetadata.R',local=knitr::knit_global())
if (!exists("year")) {
year <- 2012
}
source('LoadUSdata.R',local=knitr::knit_global())
cat(paste(year,"US IO data for states successfully loaded."))
source('Load1Rdata.R',local=knitr::knit_global())
cat(paste(year,"single-region IO data for states successfully loaded."))
source('Load2Rdata.R',local=knitr::knit_global())
cat(paste(year,"two region IO data for states successfully loaded."))
Check state IO tables
1. Check if industry output from state Make and Use are almost equal (<= 0.01).
failures <- c()
for (state in states[states != "Overseas"]) {
# Calculate Industry Output (x) from Make and Use
x_make <- rowSums(State_Summary_Make_ls[[state]])
x_use <- colSums(State_Summary_Use_ls[[state]][, rownames(US_Summary_Make)])
# Calculate relative differences in x_make and x_use
rel_diff <- (x_use - x_make)/x_make
# Compare rel_diff against 0 with a tolerance of 1E-3
failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
colnames(failures_state)[1:3] <- c("Industry", "Relative Diff", "Validation")
failures <- rbind(failures, failures_state)
}
cat("\n")
reportValidationResult(failures)
Compare state to US IO tables (negativity in the same cell & state sum == US totals)
2. Sum of each cell across all state Make tables must almost equal (<= 0.001) the same cell in US Make table.
df0 <- US_Summary_Make
df1 <- Reduce("+", State_Summary_Make_ls)
rownames(df1) <- gsub(".*\\.", "", rownames(df1))
df1 <- df1[rownames(df0), colnames(df0)]
# Compare aggregated state Make against US Make
failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E-3)[["Failure"]]
reportValidationResult(failures)
#### (DEPRECATED) 2. Check commodity output
#### (DEPRECATED) 2.1. Check if commodity output from state Make and Use are almost equal (relative difference <= 0.01).
failures <- c()
for (state in states[states != "Overseas"]) {
# Calculate Commodity Output (q) from Make and Use
q_make <- colSums(State_Summary_Make_ls[[state]])
q_use <- rowSums(State_Summary_Use_ls[[state]][rownames(US_Summary_Use), colnames(US_Summary_Use)])
# Calculate relative differences in q_make and q_use
rel_diff <- (q_use - q_make)/q_make
# Compare rel_diff against 0 with a tolerance of 1E-3
failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation")
failures <- rbind(failures, failures_state)
}
cat("\n")
reportValidationResult(failures)
#### (DEPRECATED) 2.2. Check if sum of commodity output from state Use and sum demand from Domestic Use are almost equal (relative difference <= 0.01).
q_sum <- Reduce("+", State_Summary_CommodityOutput_ls)
demand_sum <- Reduce("+", lapply(State_Summary_DomesticUse_ls, rowSums))[commodities]
# Calculate relative differences in q_sum and demand_sum
rel_diff <- (q_sum - demand_sum)/demand_sum
# Compare rel_diff against 0 with a tolerance of 1E-3
failures <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]]
colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation")
cat("\n")
reportValidationResult(failures)
3. There should not be any negative values in state Make table, unless they are negative in US Make table.
Note: only exception being Overseas, which isn't used for further calculations, and if the same cell in US Make table is also negative.
US_failures <- formatValidationResult(US_Summary_Make,
abs_diff = TRUE, tolerance = 0)[["Pass"]]
colnames(US_failures) <- c("Industry", "Commodity")
# Validate Make table and extract failures for each state
failures <- data.frame()
for (state in states[states != "Overseas"]) {
df <- as.data.frame(State_Summary_Make_ls[[state]])
# Check if there is zero in state Make table
failures_state <- formatValidationResult(df, abs_diff = TRUE, tolerance = 0)[["Pass"]]
colnames(failures_state) <- c("Industry", "Commodity")
# If failure is in US Make table, remove it from state failures
if (nrow(failures_state) > 0) {
failures_state$index <- paste(sub(".*\\.", "", failures_state$Industry),
failures_state$Commodity)
US_state_diff <- setdiff(failures_state$index,
paste(US_failures$Industry, US_failures$Commodity))
failures_state <- failures_state[failures_state$index %in% US_state_diff, ]
}
failures_state$index <- rownames(failures_state) <- NULL
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
4. Sum of each industry's output across all states must almost equal (<= 1E7, or $10 million by industry) the industry output in US Make Table.
The threshold is set to 1E7 because there are differences (within +/- $10 million) between US industry output summed from Make and that summed from Use, comparing sum of state industry output (summed from state Use) to US industry output summed from US Make should account for those inherent differences at the national level.
df0 <- as.data.frame(rowSums(US_Summary_Make))
df1 <- Reduce("+", State_Summary_IndustryOutput_ls)
rownames(df1) <- gsub(".*\\.", "", rownames(df1))
df1 <- df1[rownames(df0), ]
# Compare aggregated state industry output against US industry output
failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E7)[["Failure"]]
colnames(failures) <- c("Industry", "")
reportValidationResult(failures)
5. Sum of each commodity's output across all states must almost equal (<= 1E7, or $10 million by commodity) the commodity output in US Make Table.
The threshold is set to 1E7 because there are differences (within +/- $10 million) between US industry output summed from Make and that summed from Use, comparing sum of state industry output (summed from state Use) to US industry output summed from US Make should account for those inherent differences at the national level.
df0 <- as.data.frame(colSums(US_Summary_Make))
df1 <- Reduce("+", State_Summary_CommodityOutput_ls)
rownames(df1) <- gsub(".*\\.", "", rownames(df1))
df1 <- df1[rownames(df0), ]
# Compare aggregated state commodity output against US commodity output
failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E7)[["Failure"]]
colnames(failures) <- c("Commodity", "")
reportValidationResult(failures)
6. Sum of each commodity's output across all states must almost equal (<= 1E^7, or $10 million by commodity) commodity output in US Use Table.
Note: even if the threshold is met, track the difference for each commodity. Save result as a type of quality control check.
df0 <- as.data.frame(rowSums(US_Summary_Use))
df1 <- Reduce("+", State_Summary_CommodityOutput_ls)
rownames(df1) <- gsub(".*\\.", "", rownames(df1))
df1 <- df1[rownames(df0), ]
# Compare aggregated state commodity output against row sum of US domestic Use
failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1.11E7)[["Failure"]]
colnames(failures) <- c("Commodity", "")
reportValidationResult(failures)
diff <- df1 - df0
colnames(diff) <- "q_state_sum - q_US_use"
knitr::kable(diff, "simple")
7. All cells that are zero in US Make table must remain zero in state Make tables. Find zero values in US Make table.
US_zeros <- formatValidationResult(US_Summary_Make - 0, abs_diff = TRUE,
tolerance = 0)[["Pass"]]
colnames(US_zeros) <- c("Industry", "Commodity")
failures <- data.frame()
for (state in states) {
# Find zero values in state Make table
state_zeros <- formatValidationResult(State_Summary_Make_ls[[state]] - 0,
abs_diff = TRUE, tolerance = 0)[["Pass"]]
colnames(state_zeros)[1:2] <- c("Industry", "Commodity")
state_zeros$index <- paste(sub(".*\\.", "", state_zeros$Industry), state_zeros$Commodity)
# Find difference between state_zeros and US_zeros: zeros in US that are not zeros in state
US_state_diff <- setdiff(paste(US_zeros$Industry, US_zeros$Commodity), state_zeros$index)
failures_state <- state_zeros[state_zeros$index %in% US_state_diff, ]
failures_state$index <- rownames(failures_state) <- NULL
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
8. Sum of each cell across all state Use tables must almost equal (<= 5E6, or $5 million) the same cell in US Use table. This validates that Total state demand == Total national demand.
Note: failures associated with 'F050 - Imports' are acceptable. Because state imports are not directly derived from US imports, a gap in imports between state sum and national total is reasonable.
StateUseValidationFailures_absdiff <- validateStateUseAgainstNationlUse(domestic = FALSE, rel_diff = FALSE)
StateDomesticUseValidationFailures_absdiff <- validateStateUseAgainstNationlUse(domestic = TRUE, rel_diff = FALSE)
StateUseValidationFailures_reldiff <- validateStateUseAgainstNationlUse(domestic = FALSE, rel_diff = TRUE)
StateDomesticUseValidationFailures_reldiff <- validateStateUseAgainstNationlUse(domestic = TRUE, rel_diff = TRUE)
# State Use tables
cat("##### 8.1 State Use tables (checking absolute differences)\n")
reportValidationResult(StateUseValidationFailures_absdiff)
# State Domestic Use tables
cat("##### 8.2 State Domestic Use tables (checking absolute differences)\n")
reportValidationResult(StateDomesticUseValidationFailures_absdiff)
# State Use tables
cat("##### 8.3 State Use tables (checking relative differences)\n")
reportValidationResult(StateUseValidationFailures_reldiff)
# State Domestic Use tables
cat("##### 8.4 State Domestic Use tables (checking relative differences)\n")
reportValidationResult(StateDomesticUseValidationFailures_reldiff)
Check two-region model results
9. Check if commodity output from two-region Make and Domestic Use are almost equal (relative difference <= 0.01).
failures <- c()
for (state in states[states != "Overseas"]) {
# Calculate Commodity Output (q) from Make and Use
q_make <- colSums(TwoRegionMake_ls[[state]])
use_2r <- TwoRegionDomesticUsewithTrade_ls[[state]]
q_use <- c(rowSums(use_2r[["SoI2SoI"]][, c(industries, FD_cols, ITA_col, "ExportResidual")]) +
rowSums(use_2r[["SoI2RoUS"]][, c(industries, FD_cols, ITA_col)]),
rowSums(use_2r[["RoUS2RoUS"]][, c(industries, FD_cols, ITA_col, "ExportResidual")]) +
rowSums(use_2r[["RoUS2SoI"]][, c(industries, FD_cols, ITA_col)]))
# Calculate relative differences in q_make and q_use
rel_diff <- (q_use - q_make)/q_make
# Compare rel_diff against 0 with a tolerance of 1E-3
failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation")
failures <- rbind(failures, failures_state)
}
cat("\n")
reportValidationResult(failures)
10. If SoI commodity output == 0, SoI2SoI ICF ratio == 0
failures <- c()
for (state in states[states != "Overseas"]) {
# Find zero values in SoI commodity output
CO_zeros <- formatValidationResult(State_Summary_CommodityOutput_ls[[state]] - 0,
abs_diff = TRUE, tolerance = 0)[["Pass"]]
colnames(CO_zeros) <- c("Commodity", "")
# Find zero values in SoI2SoI ICF ratio
ICF <- generateDomestic2RegionICFs(state, year, ioschema = 2012, iolevel = "Summary",
ICF_sensitivity_analysis = FALSE, adjust_by = 0)
SoI2SoI_ICF_zeros <- ICF[ICF$SoI2SoI == 0, 1]
# Find difference between SoI2SoI_ICF_zeros and CO_zeros:
# zeros in CO that are not zeros in SoI2SoI_ICF
diff <- setdiff(CO_zeros$Commodity, SoI2SoI_ICF_zeros)
failures_state <- SoI2SoI_ICF_zeros[SoI2SoI_ICF_zeros %in% diff]
failures <- c(failures, failures_state)
}
cat("\n")
reportValidationResult(failures)
11. SoI and RoUS interregional exports >= 0, interregional imports >= 0
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
df <- cbind(TwoRegionTable_state[["SoI2SoI"]][, c("InterregionalImports", "InterregionalExports")],
TwoRegionTable_state[["RoUS2RoUS"]][, c("InterregionalImports", "InterregionalExports")])
df <- as.data.frame(sapply(df, round, 1))
rownames(df) <- rownames(TwoRegionTable_state[["SoI2SoI"]])
colnames(df) <- paste(rep(paste(state, c("SoI2SoI", "RoUS2RoUS"), sep = "_"), each = 2),
colnames(df), sep = "$")
# Compare SoI and RoUS interregional exports and imports against 0
failures_state <- formatValidationResult(df, abs_diff = FALSE, tolerance = 0)[["Pass"]]
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
12. SoI net exports + RoUS net exports == 0
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
df <- TwoRegionTable_state[["SoI2SoI"]][, "NetExports", drop = FALSE] +
TwoRegionTable_state[["RoUS2RoUS"]][, "NetExports", drop = FALSE]
df <- as.data.frame(sapply(df, round, 1))
rownames(df) <- rownames(TwoRegionTable_state[["SoI2SoI"]])
# Compare SoI net exports + RoUS net exports against 0
failures_state <- formatValidationResult(df, abs_diff = TRUE, tolerance = 0)[["Pass"]]
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
13. Check row sum of SoI2SoI <= state commodity supply. Row sum of RoUS2RoUS <= RoUS commodity supply.
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare state commodity supply
df0 <- cbind.data.frame(State_Summary_CommodityOutput_ls[[state]][, "Output", drop = FALSE],
colSums(US_Summary_Make))
colnames(df0) <- c("StateCommOutput", "USCommOutput")
df0$RoUSCommOutput <- df0$USCommOutput - df0$StateCommOutput
df0$USCommOutput <- NULL
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary"), "ExportResidual")
df1 <- cbind.data.frame(rowSums(TwoRegionTable_state[["SoI2SoI"]][, columns]),
rowSums(TwoRegionTable_state[["RoUS2RoUS"]][, columns]))
colnames(df1) <- c(state, paste0(state, "'s RoUS"))
# Compare row sum of SoI2SoI against state commodity supply
failures_state <- formatValidationResult(df1 - df0, abs_diff = FALSE,
tolerance = 1E7)[["Failure"]]
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
14. Value in SoI2SoI and RoUS2RoUS can be negative only when the same cell is negative in national Use table
# Find negative values in US Use table
US_negatives <- formatValidationResult(US_Summary_DomesticUse,
abs_diff = FALSE, tolerance = 0)[["Pass"]]
colnames(US_negatives) <- c("Commodity", "Industry")
# Validate the position of zero values in state Make tables
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary"))
# Find negative values in SoI2SoI Use table
df_SoI <- TwoRegionTable_state[["SoI2SoI"]][, columns]
SoI_negatives <- formatValidationResult(df_SoI, abs_diff = FALSE,
tolerance = 0)[["Pass"]]
colnames(SoI_negatives)[1:2] <- c("Commodity", "Industry")
SoI_negatives$index <- paste(SoI_negatives$Commodity, SoI_negatives$Industry)
SoI_negatives$table <- paste(state, "SoI2SoI Use")
# Find difference between SoI_negatives and US_negatives:
# negatives in SoI2SoI that are not negatives in US Use
US_SoI_diff <- setdiff(SoI_negatives$index,
paste(US_negatives$Commodity, US_negatives$Industry))
# Find negative values in RoUS2RoUS Use table
df_RoUS <- TwoRegionTable_state[["RoUS2RoUS"]][, columns]
RoUS_negatives <- formatValidationResult(df_RoUS, abs_diff = FALSE,
tolerance = 0)[["Pass"]]
colnames(RoUS_negatives)[1:2] <- c("Commodity", "Industry")
RoUS_negatives$index <- paste(RoUS_negatives$Commodity, RoUS_negatives$Industry)
RoUS_negatives$table <- paste(state, "RoUS2RoUS Use")
# Find difference between RoUS_negatives and US_negatives:
# negatives in RoUS2RoUS that are not negatives in US Use
US_RoUS_diff <- setdiff(RoUS_negatives$index,
paste(US_negatives$Commodity, US_negatives$Industry))
# Compile failures
failures_state <- rbind(SoI_negatives[SoI_negatives$index %in% US_SoI_diff, ],
RoUS_negatives[RoUS_negatives$index %in% US_RoUS_diff, ])
failures_state$index <- rownames(failures_state) <- NULL
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
15. SoI interregional imports == RoUS interregional exports, or difference <= 0.001
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
# Prepare df0 and df1
df0 <- TwoRegionTable_state[["SoI2SoI"]][, "InterregionalImports", drop = FALSE]
df1 <- TwoRegionTable_state[["RoUS2RoUS"]][, "InterregionalExports", drop = FALSE]
# Compare SoI interregional imports against RoUS interregional exports
failures_state <- formatValidationResult(df0 - df1, abs_diff = FALSE,
tolerance = 1E-3)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
16.1. Total state commodity supply == state demand by intermediate consumption, plus final demand (except imports and international trade adjustment) + Interregional Exports + Export Residual. Difference must be <= 0.001.
failures <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
columns <- c(getVectorOfCodes("Summary", "Industry"),
setdiff(getFinalDemandCodes("Summary"),
getVectorOfCodes("Summary", "Import")),
"F051", "InterregionalExports", "ExportResidual")
# Prepare df0 and df1
df0 <- State_Summary_CommodityOutput_ls[[state]][, "Output", drop = FALSE]
df1 <- as.data.frame(rowSums(TwoRegionTable_state[["SoI2SoI"]][, columns]))
# Compare state commodity supply against state demand by intermediate consumption,
# plus final demand (except imports) + Interregional Exports
failures_state <- formatValidationResult(df0 - df1, abs_diff = TRUE,
tolerance = 1E-3)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
16.2. Total SoI and RoUS commodity supply (output) == SoI and RoUS demand (domestic intermediate consumption + ITA + Export Residual).
failures <- c()
for (state in states[states != "Overseas"]) {
# Calculate Commodity Output (q) from Make and Use
q <- TwoRegionCommodityOutput_ls[[state]]
q_2r_use <- rowSums(TwoRegionDomesticUse_ls[[state]]) + TwoRegionITA_ls[[state]] + c(TwoRegionDomesticUsewithTrade_ls[[state]][["SoI2SoI"]][, "ExportResidual"], TwoRegionDomesticUsewithTrade_ls[[state]][["RoUS2RoUS"]][, "ExportResidual"])
# Calculate relative differences in q_make and q_use
rel_diff <- (q - q_2r_use[names(q)])/q_2r_use[names(q)]
# Compare rel_diff against 0 with a tolerance of 1E-3
failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]]
if (nrow(failures_state) > 0) {
failures_state$State <- state
}
colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation")
failures <- rbind(failures, failures_state)
}
cat("\n")
reportValidationResult(failures)
17. Number of negative cells in SoI2SoI, SoI2RoUS, RoUS2SoI and RoUS2RoUS <= Number of negative cells in national Use table
US_negatives <- formatValidationResult(US_Summary_DomesticUse,
abs_diff = FALSE, tolerance = 0)[["Pass"]]
colnames(US_negatives) <- c("Commodity", "Industry")
failure <- data.frame()
for (state in states[states != "Overseas"]) {
# Prepare domestic 2-region Use tables
TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]]
columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary"))
failures_state <- data.frame()
for (table in names(TwoRegionTable_state)[1:4]) {
# Find number of negative cells in two-region Use table
df <- TwoRegionTable_state[[table]][, columns]
state_negatives <- formatValidationResult(df, abs_diff = FALSE, tolerance = 0)[["Pass"]]
colnames(state_negatives) <- c("Commodity", "Industry")
state_negatives$index <- paste(state_negatives$Commodity, state_negatives$Industry)
state_negatives$table <- paste(state, table, "Use")
# Find difference between state_negatives and US_negatives:
# negatives in RoUS2RoUS that are not negatives in US Use
US_state_diff <- setdiff(state_negatives$index,
paste(US_negatives$Commodity, US_negatives$Industry))
# Compile failures
failures_table <- state_negatives[state_negatives$index %in% US_state_diff, ]
failures_table$index <- NULL
failures_table <- rbind(failures_state, failures_table)
}
failures <- rbind(failures, failures_state)
}
reportValidationResult(failures)
18. Non-square model verification. Validate L matrix of two-region model and final demand against SoI and RoUS output.
Absolute difference: L*y - output <= 1^6, or $1 million.
Relative difference: (L*y - output)/output <= 1^-2, or 1%.
Failures_ls <- list()
Failures_RelDiff_ls <- list()
for (state in state.name) {
# Generate validation results
SoI_2r_LagaintsOutput_Validation <- validateTwoRegionLagainstOutput(state, year, ioschema = 2012, iolevel = "Summary")
# Failures
Failures_ls[[state]] <- formatValidationResult(SoI_2r_LagaintsOutput_Validation$Validation[, "L*y-output", drop = FALSE],
abs_diff = TRUE, tolerance = 1E6)[["Failure"]]
Failures_RelDiff_ls[[state]] <- formatValidationResult(SoI_2r_LagaintsOutput_Validation$Validation[, "rel_diff", drop = FALSE],
abs_diff = TRUE, tolerance = 1E-2)[["Failure"]]
}
for (state in state.name) {
cat(paste0("##### 18.", which(states == state)), state, "and Rest of the US\n")
cat("Absolute Difference: \n")
reportValidationResult(Failures_ls[[state]])
cat("Relative Difference: \n")
reportValidationResult(Failures_RelDiff_ls[[state]])
}
USEPA/stateio documentation built on Feb. 12, 2024, 6:41 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
This document presents validation results of r year summary-level state IO model.
knitr::opts_chunk$set(echo = FALSE, message = FALSE, results = "asis") logging::basicConfig(level = 40)
library(devtools) load_all()
Prepare data
0. Load data
source('DefineMetadata.R',local=knitr::knit_global())
if (!exists("year")) { year <- 2012 }
source('LoadUSdata.R',local=knitr::knit_global()) cat(paste(year,"US IO data for states successfully loaded."))
source('Load1Rdata.R',local=knitr::knit_global()) cat(paste(year,"single-region IO data for states successfully loaded."))
source('Load2Rdata.R',local=knitr::knit_global()) cat(paste(year,"two region IO data for states successfully loaded."))
Check state IO tables
1. Check if industry output from state Make and Use are almost equal (<= 0.01).
failures <- c() for (state in states[states != "Overseas"]) { # Calculate Industry Output (x) from Make and Use x_make <- rowSums(State_Summary_Make_ls[[state]]) x_use <- colSums(State_Summary_Use_ls[[state]][, rownames(US_Summary_Make)]) # Calculate relative differences in x_make and x_use rel_diff <- (x_use - x_make)/x_make # Compare rel_diff against 0 with a tolerance of 1E-3 failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } colnames(failures_state)[1:3] <- c("Industry", "Relative Diff", "Validation") failures <- rbind(failures, failures_state) } cat("\n") reportValidationResult(failures)
Compare state to US IO tables (negativity in the same cell & state sum == US totals)
2. Sum of each cell across all state Make tables must almost equal (<= 0.001) the same cell in US Make table.
df0 <- US_Summary_Make df1 <- Reduce("+", State_Summary_Make_ls) rownames(df1) <- gsub(".*\\.", "", rownames(df1)) df1 <- df1[rownames(df0), colnames(df0)] # Compare aggregated state Make against US Make failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E-3)[["Failure"]] reportValidationResult(failures)
#### (DEPRECATED) 2. Check commodity output #### (DEPRECATED) 2.1. Check if commodity output from state Make and Use are almost equal (relative difference <= 0.01). failures <- c() for (state in states[states != "Overseas"]) { # Calculate Commodity Output (q) from Make and Use q_make <- colSums(State_Summary_Make_ls[[state]]) q_use <- rowSums(State_Summary_Use_ls[[state]][rownames(US_Summary_Use), colnames(US_Summary_Use)]) # Calculate relative differences in q_make and q_use rel_diff <- (q_use - q_make)/q_make # Compare rel_diff against 0 with a tolerance of 1E-3 failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation") failures <- rbind(failures, failures_state) } cat("\n") reportValidationResult(failures)
#### (DEPRECATED) 2.2. Check if sum of commodity output from state Use and sum demand from Domestic Use are almost equal (relative difference <= 0.01). q_sum <- Reduce("+", State_Summary_CommodityOutput_ls) demand_sum <- Reduce("+", lapply(State_Summary_DomesticUse_ls, rowSums))[commodities] # Calculate relative differences in q_sum and demand_sum rel_diff <- (q_sum - demand_sum)/demand_sum # Compare rel_diff against 0 with a tolerance of 1E-3 failures <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]] colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation") cat("\n") reportValidationResult(failures)
3. There should not be any negative values in state Make table, unless they are negative in US Make table.
Note: only exception being Overseas, which isn't used for further calculations, and if the same cell in US Make table is also negative.
US_failures <- formatValidationResult(US_Summary_Make, abs_diff = TRUE, tolerance = 0)[["Pass"]] colnames(US_failures) <- c("Industry", "Commodity") # Validate Make table and extract failures for each state failures <- data.frame() for (state in states[states != "Overseas"]) { df <- as.data.frame(State_Summary_Make_ls[[state]]) # Check if there is zero in state Make table failures_state <- formatValidationResult(df, abs_diff = TRUE, tolerance = 0)[["Pass"]] colnames(failures_state) <- c("Industry", "Commodity") # If failure is in US Make table, remove it from state failures if (nrow(failures_state) > 0) { failures_state$index <- paste(sub(".*\\.", "", failures_state$Industry), failures_state$Commodity) US_state_diff <- setdiff(failures_state$index, paste(US_failures$Industry, US_failures$Commodity)) failures_state <- failures_state[failures_state$index %in% US_state_diff, ] } failures_state$index <- rownames(failures_state) <- NULL failures <- rbind(failures, failures_state) } reportValidationResult(failures)
4. Sum of each industry's output across all states must almost equal (<= 1E7, or $10 million by industry) the industry output in US Make Table.
The threshold is set to 1E7 because there are differences (within +/- $10 million) between US industry output summed from Make and that summed from Use, comparing sum of state industry output (summed from state Use) to US industry output summed from US Make should account for those inherent differences at the national level.
df0 <- as.data.frame(rowSums(US_Summary_Make)) df1 <- Reduce("+", State_Summary_IndustryOutput_ls) rownames(df1) <- gsub(".*\\.", "", rownames(df1)) df1 <- df1[rownames(df0), ] # Compare aggregated state industry output against US industry output failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E7)[["Failure"]] colnames(failures) <- c("Industry", "") reportValidationResult(failures)
5. Sum of each commodity's output across all states must almost equal (<= 1E7, or $10 million by commodity) the commodity output in US Make Table.
The threshold is set to 1E7 because there are differences (within +/- $10 million) between US industry output summed from Make and that summed from Use, comparing sum of state industry output (summed from state Use) to US industry output summed from US Make should account for those inherent differences at the national level.
df0 <- as.data.frame(colSums(US_Summary_Make)) df1 <- Reduce("+", State_Summary_CommodityOutput_ls) rownames(df1) <- gsub(".*\\.", "", rownames(df1)) df1 <- df1[rownames(df0), ] # Compare aggregated state commodity output against US commodity output failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1E7)[["Failure"]] colnames(failures) <- c("Commodity", "") reportValidationResult(failures)
6. Sum of each commodity's output across all states must almost equal (<= 1E^7, or $10 million by commodity) commodity output in US Use Table.
Note: even if the threshold is met, track the difference for each commodity. Save result as a type of quality control check.
df0 <- as.data.frame(rowSums(US_Summary_Use)) df1 <- Reduce("+", State_Summary_CommodityOutput_ls) rownames(df1) <- gsub(".*\\.", "", rownames(df1)) df1 <- df1[rownames(df0), ] # Compare aggregated state commodity output against row sum of US domestic Use failures <- formatValidationResult(df1 - df0, abs_diff = TRUE, tolerance = 1.11E7)[["Failure"]] colnames(failures) <- c("Commodity", "") reportValidationResult(failures) diff <- df1 - df0 colnames(diff) <- "q_state_sum - q_US_use" knitr::kable(diff, "simple")
7. All cells that are zero in US Make table must remain zero in state Make tables. Find zero values in US Make table.
US_zeros <- formatValidationResult(US_Summary_Make - 0, abs_diff = TRUE, tolerance = 0)[["Pass"]] colnames(US_zeros) <- c("Industry", "Commodity") failures <- data.frame() for (state in states) { # Find zero values in state Make table state_zeros <- formatValidationResult(State_Summary_Make_ls[[state]] - 0, abs_diff = TRUE, tolerance = 0)[["Pass"]] colnames(state_zeros)[1:2] <- c("Industry", "Commodity") state_zeros$index <- paste(sub(".*\\.", "", state_zeros$Industry), state_zeros$Commodity) # Find difference between state_zeros and US_zeros: zeros in US that are not zeros in state US_state_diff <- setdiff(paste(US_zeros$Industry, US_zeros$Commodity), state_zeros$index) failures_state <- state_zeros[state_zeros$index %in% US_state_diff, ] failures_state$index <- rownames(failures_state) <- NULL failures <- rbind(failures, failures_state) } reportValidationResult(failures)
8. Sum of each cell across all state Use tables must almost equal (<= 5E6, or $5 million) the same cell in US Use table. This validates that Total state demand == Total national demand.
Note: failures associated with 'F050 - Imports' are acceptable. Because state imports are not directly derived from US imports, a gap in imports between state sum and national total is reasonable.
StateUseValidationFailures_absdiff <- validateStateUseAgainstNationlUse(domestic = FALSE, rel_diff = FALSE) StateDomesticUseValidationFailures_absdiff <- validateStateUseAgainstNationlUse(domestic = TRUE, rel_diff = FALSE) StateUseValidationFailures_reldiff <- validateStateUseAgainstNationlUse(domestic = FALSE, rel_diff = TRUE) StateDomesticUseValidationFailures_reldiff <- validateStateUseAgainstNationlUse(domestic = TRUE, rel_diff = TRUE) # State Use tables cat("##### 8.1 State Use tables (checking absolute differences)\n") reportValidationResult(StateUseValidationFailures_absdiff) # State Domestic Use tables cat("##### 8.2 State Domestic Use tables (checking absolute differences)\n") reportValidationResult(StateDomesticUseValidationFailures_absdiff) # State Use tables cat("##### 8.3 State Use tables (checking relative differences)\n") reportValidationResult(StateUseValidationFailures_reldiff) # State Domestic Use tables cat("##### 8.4 State Domestic Use tables (checking relative differences)\n") reportValidationResult(StateDomesticUseValidationFailures_reldiff)
Check two-region model results
9. Check if commodity output from two-region Make and Domestic Use are almost equal (relative difference <= 0.01).
failures <- c() for (state in states[states != "Overseas"]) { # Calculate Commodity Output (q) from Make and Use q_make <- colSums(TwoRegionMake_ls[[state]]) use_2r <- TwoRegionDomesticUsewithTrade_ls[[state]] q_use <- c(rowSums(use_2r[["SoI2SoI"]][, c(industries, FD_cols, ITA_col, "ExportResidual")]) + rowSums(use_2r[["SoI2RoUS"]][, c(industries, FD_cols, ITA_col)]), rowSums(use_2r[["RoUS2RoUS"]][, c(industries, FD_cols, ITA_col, "ExportResidual")]) + rowSums(use_2r[["RoUS2SoI"]][, c(industries, FD_cols, ITA_col)])) # Calculate relative differences in q_make and q_use rel_diff <- (q_use - q_make)/q_make # Compare rel_diff against 0 with a tolerance of 1E-3 failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation") failures <- rbind(failures, failures_state) } cat("\n") reportValidationResult(failures)
10. If SoI commodity output == 0, SoI2SoI ICF ratio == 0
failures <- c() for (state in states[states != "Overseas"]) { # Find zero values in SoI commodity output CO_zeros <- formatValidationResult(State_Summary_CommodityOutput_ls[[state]] - 0, abs_diff = TRUE, tolerance = 0)[["Pass"]] colnames(CO_zeros) <- c("Commodity", "") # Find zero values in SoI2SoI ICF ratio ICF <- generateDomestic2RegionICFs(state, year, ioschema = 2012, iolevel = "Summary", ICF_sensitivity_analysis = FALSE, adjust_by = 0) SoI2SoI_ICF_zeros <- ICF[ICF$SoI2SoI == 0, 1] # Find difference between SoI2SoI_ICF_zeros and CO_zeros: # zeros in CO that are not zeros in SoI2SoI_ICF diff <- setdiff(CO_zeros$Commodity, SoI2SoI_ICF_zeros) failures_state <- SoI2SoI_ICF_zeros[SoI2SoI_ICF_zeros %in% diff] failures <- c(failures, failures_state) } cat("\n") reportValidationResult(failures)
11. SoI and RoUS interregional exports >= 0, interregional imports >= 0
failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] df <- cbind(TwoRegionTable_state[["SoI2SoI"]][, c("InterregionalImports", "InterregionalExports")], TwoRegionTable_state[["RoUS2RoUS"]][, c("InterregionalImports", "InterregionalExports")]) df <- as.data.frame(sapply(df, round, 1)) rownames(df) <- rownames(TwoRegionTable_state[["SoI2SoI"]]) colnames(df) <- paste(rep(paste(state, c("SoI2SoI", "RoUS2RoUS"), sep = "_"), each = 2), colnames(df), sep = "$") # Compare SoI and RoUS interregional exports and imports against 0 failures_state <- formatValidationResult(df, abs_diff = FALSE, tolerance = 0)[["Pass"]] failures <- rbind(failures, failures_state) } reportValidationResult(failures)
12. SoI net exports + RoUS net exports == 0
failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] df <- TwoRegionTable_state[["SoI2SoI"]][, "NetExports", drop = FALSE] + TwoRegionTable_state[["RoUS2RoUS"]][, "NetExports", drop = FALSE] df <- as.data.frame(sapply(df, round, 1)) rownames(df) <- rownames(TwoRegionTable_state[["SoI2SoI"]]) # Compare SoI net exports + RoUS net exports against 0 failures_state <- formatValidationResult(df, abs_diff = TRUE, tolerance = 0)[["Pass"]] failures <- rbind(failures, failures_state) } reportValidationResult(failures)
13. Check row sum of SoI2SoI <= state commodity supply. Row sum of RoUS2RoUS <= RoUS commodity supply.
failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare state commodity supply df0 <- cbind.data.frame(State_Summary_CommodityOutput_ls[[state]][, "Output", drop = FALSE], colSums(US_Summary_Make)) colnames(df0) <- c("StateCommOutput", "USCommOutput") df0$RoUSCommOutput <- df0$USCommOutput - df0$StateCommOutput df0$USCommOutput <- NULL # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary"), "ExportResidual") df1 <- cbind.data.frame(rowSums(TwoRegionTable_state[["SoI2SoI"]][, columns]), rowSums(TwoRegionTable_state[["RoUS2RoUS"]][, columns])) colnames(df1) <- c(state, paste0(state, "'s RoUS")) # Compare row sum of SoI2SoI against state commodity supply failures_state <- formatValidationResult(df1 - df0, abs_diff = FALSE, tolerance = 1E7)[["Failure"]] failures <- rbind(failures, failures_state) } reportValidationResult(failures)
14. Value in SoI2SoI and RoUS2RoUS can be negative only when the same cell is negative in national Use table
# Find negative values in US Use table US_negatives <- formatValidationResult(US_Summary_DomesticUse, abs_diff = FALSE, tolerance = 0)[["Pass"]] colnames(US_negatives) <- c("Commodity", "Industry") # Validate the position of zero values in state Make tables failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary")) # Find negative values in SoI2SoI Use table df_SoI <- TwoRegionTable_state[["SoI2SoI"]][, columns] SoI_negatives <- formatValidationResult(df_SoI, abs_diff = FALSE, tolerance = 0)[["Pass"]] colnames(SoI_negatives)[1:2] <- c("Commodity", "Industry") SoI_negatives$index <- paste(SoI_negatives$Commodity, SoI_negatives$Industry) SoI_negatives$table <- paste(state, "SoI2SoI Use") # Find difference between SoI_negatives and US_negatives: # negatives in SoI2SoI that are not negatives in US Use US_SoI_diff <- setdiff(SoI_negatives$index, paste(US_negatives$Commodity, US_negatives$Industry)) # Find negative values in RoUS2RoUS Use table df_RoUS <- TwoRegionTable_state[["RoUS2RoUS"]][, columns] RoUS_negatives <- formatValidationResult(df_RoUS, abs_diff = FALSE, tolerance = 0)[["Pass"]] colnames(RoUS_negatives)[1:2] <- c("Commodity", "Industry") RoUS_negatives$index <- paste(RoUS_negatives$Commodity, RoUS_negatives$Industry) RoUS_negatives$table <- paste(state, "RoUS2RoUS Use") # Find difference between RoUS_negatives and US_negatives: # negatives in RoUS2RoUS that are not negatives in US Use US_RoUS_diff <- setdiff(RoUS_negatives$index, paste(US_negatives$Commodity, US_negatives$Industry)) # Compile failures failures_state <- rbind(SoI_negatives[SoI_negatives$index %in% US_SoI_diff, ], RoUS_negatives[RoUS_negatives$index %in% US_RoUS_diff, ]) failures_state$index <- rownames(failures_state) <- NULL failures <- rbind(failures, failures_state) } reportValidationResult(failures)
15. SoI interregional imports == RoUS interregional exports, or difference <= 0.001
failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] # Prepare df0 and df1 df0 <- TwoRegionTable_state[["SoI2SoI"]][, "InterregionalImports", drop = FALSE] df1 <- TwoRegionTable_state[["RoUS2RoUS"]][, "InterregionalExports", drop = FALSE] # Compare SoI interregional imports against RoUS interregional exports failures_state <- formatValidationResult(df0 - df1, abs_diff = FALSE, tolerance = 1E-3)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } failures <- rbind(failures, failures_state) } reportValidationResult(failures)
16.1. Total state commodity supply == state demand by intermediate consumption, plus final demand (except imports and international trade adjustment) + Interregional Exports + Export Residual. Difference must be <= 0.001.
failures <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] columns <- c(getVectorOfCodes("Summary", "Industry"), setdiff(getFinalDemandCodes("Summary"), getVectorOfCodes("Summary", "Import")), "F051", "InterregionalExports", "ExportResidual") # Prepare df0 and df1 df0 <- State_Summary_CommodityOutput_ls[[state]][, "Output", drop = FALSE] df1 <- as.data.frame(rowSums(TwoRegionTable_state[["SoI2SoI"]][, columns])) # Compare state commodity supply against state demand by intermediate consumption, # plus final demand (except imports) + Interregional Exports failures_state <- formatValidationResult(df0 - df1, abs_diff = TRUE, tolerance = 1E-3)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } failures <- rbind(failures, failures_state) } reportValidationResult(failures)
16.2. Total SoI and RoUS commodity supply (output) == SoI and RoUS demand (domestic intermediate consumption + ITA + Export Residual).
failures <- c() for (state in states[states != "Overseas"]) { # Calculate Commodity Output (q) from Make and Use q <- TwoRegionCommodityOutput_ls[[state]] q_2r_use <- rowSums(TwoRegionDomesticUse_ls[[state]]) + TwoRegionITA_ls[[state]] + c(TwoRegionDomesticUsewithTrade_ls[[state]][["SoI2SoI"]][, "ExportResidual"], TwoRegionDomesticUsewithTrade_ls[[state]][["RoUS2RoUS"]][, "ExportResidual"]) # Calculate relative differences in q_make and q_use rel_diff <- (q - q_2r_use[names(q)])/q_2r_use[names(q)] # Compare rel_diff against 0 with a tolerance of 1E-3 failures_state <- formatValidationResult(rel_diff, abs_diff = TRUE, tolerance = 0.01)[["Failure"]] if (nrow(failures_state) > 0) { failures_state$State <- state } colnames(failures_state)[1:3] <- c("Commodity", "Relative Diff", "Validation") failures <- rbind(failures, failures_state) } cat("\n") reportValidationResult(failures)
17. Number of negative cells in SoI2SoI, SoI2RoUS, RoUS2SoI and RoUS2RoUS <= Number of negative cells in national Use table
US_negatives <- formatValidationResult(US_Summary_DomesticUse, abs_diff = FALSE, tolerance = 0)[["Pass"]] colnames(US_negatives) <- c("Commodity", "Industry") failure <- data.frame() for (state in states[states != "Overseas"]) { # Prepare domestic 2-region Use tables TwoRegionTable_state <- TwoRegionDomesticUsewithTrade_ls[[state]] columns <- c(getVectorOfCodes("Summary", "Industry"), getFinalDemandCodes("Summary")) failures_state <- data.frame() for (table in names(TwoRegionTable_state)[1:4]) { # Find number of negative cells in two-region Use table df <- TwoRegionTable_state[[table]][, columns] state_negatives <- formatValidationResult(df, abs_diff = FALSE, tolerance = 0)[["Pass"]] colnames(state_negatives) <- c("Commodity", "Industry") state_negatives$index <- paste(state_negatives$Commodity, state_negatives$Industry) state_negatives$table <- paste(state, table, "Use") # Find difference between state_negatives and US_negatives: # negatives in RoUS2RoUS that are not negatives in US Use US_state_diff <- setdiff(state_negatives$index, paste(US_negatives$Commodity, US_negatives$Industry)) # Compile failures failures_table <- state_negatives[state_negatives$index %in% US_state_diff, ] failures_table$index <- NULL failures_table <- rbind(failures_state, failures_table) } failures <- rbind(failures, failures_state) } reportValidationResult(failures)
18. Non-square model verification. Validate L matrix of two-region model and final demand against SoI and RoUS output.
Absolute difference: L*y - output <= 1^6, or $1 million.
Relative difference: (L*y - output)/output <= 1^-2, or 1%.
Failures_ls <- list() Failures_RelDiff_ls <- list() for (state in state.name) { # Generate validation results SoI_2r_LagaintsOutput_Validation <- validateTwoRegionLagainstOutput(state, year, ioschema = 2012, iolevel = "Summary") # Failures Failures_ls[[state]] <- formatValidationResult(SoI_2r_LagaintsOutput_Validation$Validation[, "L*y-output", drop = FALSE], abs_diff = TRUE, tolerance = 1E6)[["Failure"]] Failures_RelDiff_ls[[state]] <- formatValidationResult(SoI_2r_LagaintsOutput_Validation$Validation[, "rel_diff", drop = FALSE], abs_diff = TRUE, tolerance = 1E-2)[["Failure"]] }
for (state in state.name) { cat(paste0("##### 18.", which(states == state)), state, "and Rest of the US\n") cat("Absolute Difference: \n") reportValidationResult(Failures_ls[[state]]) cat("Relative Difference: \n") reportValidationResult(Failures_RelDiff_ls[[state]]) }
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.