Training assessment

This page is a knowledge crystal of subtype assessment. The page identifier is Op_en5677
Moderator:Jouni (see all)
This page is a stub. You may improve it into a full page.
Upload data {{#opasnet_base_link:Op_en5677}}

This is a training assessment about an imaginary, simple case. The purpose is to illustrate assessment functionalities.

Scope

Question

What decisions are worth implementing in the training assessment?

Boundaries

• Time: Year 2012 - 2020

Scenarios

• Factory can reduce emissions, or continue business as usual.
• School can increase health education, decrease it to save money, or continue business as usual.

Intended users

• Anyone who wants to learn to make open assessments.

Participants

• Main participants:
  • YMAL,
  • Summer workers of YMAL in 2012,
  • Participants of [[Decision analysis and risk management 2013}}

Answer

Results

Not yet available.

Rationale

The causal diagram for the training assessment.

Assessment-specific data

Decisions

Decisions(-)

Obs	Decisionmaker	Decision	Option	Variable	Cell	Change	Unit	Amount	Description
1	Factory	Cleaning.policy	Reduce emissions	exposure	Year:2020	Multiply	-	0.5
2	School	Health.promotion	Increase health education	health.impact	Year:2020	Multiply	-	0.9
3	School	Health.promotion	Promotion budget reduced	health.impact	Year:2020	Multiply	-	1.1

Probabilities

Probabilities(P)

Obs	Stakeholder	Variable	Cell	Probability	Description
1	City of Kuopio	exposure	Cleaning.policy: Reduce emissions	0.8
2	City of Kuopio	exposure	Cleaning.policy: BAU	0.2
3	Factoy	health.impact	Health.promotion: Increase health education	0.1
4	Factoy	health.impact	Health.promotion: Promotion budget reduced	0.4
5	Factoy	health.impact	Health.promotion: BAU	0.5

Endpoints

Endpoints(-)

Obs	Stakeholder	Variable	Cell	Model	Result	Description
1	City of Kuopio	health.impact	Year:2012	Weighted sum	1000
2	City of Kuopio	health.impact	Year:2020	Weighted sum	1000
3	Citizens	health.impact	Year:2012	Weighted sum	1000
4	Citizens	health.impact	Year:2020	Weighted sum	2000	Future years are twice as important.

Variables

• exposure: Training exposure
• health.impact: Training health impact
• training.costs: Training costs

Analyses

• Decision analysis on each policy: Which option minimises the health risks?
• Value of information (VOI) analysis for each policy about the major variables in the model and the total VOI.

Calculations

+ Show code- Hide code dectable <- data.frame( Stakeholder = rep(c("School", "City of Kuopio"), 4), Variable = rep(c("health.impact", "exposure"), each = 2), Cell = c("Year:2012,2014,2020", "Year:2012"), Result = 2 ) health.impact <- data.frame(Year = 2011:2010, Result = 1:10) exposure <- data.frame(Year = 2012, Result = 100:110) # temp <- tempovar@output[tempovar@output$health.impactSource == "Formula", ] dectable <- dectable[dectable$Stakeholder == "School" & dectable$Variable == "health.impact", ] ##################### ##### SelectCond takes conditions and slices a data.frame leaving rows that fulfil the conditions. ##### SelectCond can be used for ovariables as well, then ovariable@output is taken as the data.frame. ##### Parameters: conditions: a data.frame with conditions for slicing. It must contain columns Variable and Cell. ###### Result and Stakeholder are optional but they are typically needed for further use. ###### Stakeholder and Variable columns must have the same values at each row, ###### i.e. the Stakeholder-Variable pair must be unique. ###### There must exist a data.frame or ovariable with the name found in the Variable column. SelectCond <- function(conditions, ...) { temp <- get(as.character(dectable[1, "Variable"])) if(class(temp) == "ovariable") {temp <- temp@output} for (j in 1:nrow(dectable)) { # In the decision table format conditions are given in the "Cell"-column separated by ";". sel1 <- strsplit(as.character(dectable[j, "Cell"]), split = ";")[[1]] # ":" defines index - location matches as a condition. sel2 <- strsplit(sel1, split = ":") # No need for lapply, since strsplit is a vectorized function and current list depth is 1. # Create a list of conditions which the decision and option specific condition vector consists of. for (k in 1:length(sel1)) { # For each condition separated by ";" if (length(sel2[[k]]) > 1) { # If ":" has been used for condition k locs <- strsplit(sel2[[k]][2], split = ",")[[1]] # Split by "," for multiple locs per given index temp <- temp[temp[, sel2[[k]][1]] %in% locs , ] # Match our data.frame to the condition temp <- temp[, colnames(temp) != sel2[[k]][1] ] # Remove all indices that were used in selecting rows, because otherwise they cannot be merged. cat(j, k, "\n") } } } return(temp) } SelectCond(dectable)

+ Show code- Hide code library(OpasnetUtils) library(xtable) library(ggplot2) # List of decisions to be included in the ovariables as scenarios. decisions <- tidy(opbase.data("Op_en5677.decisions")) # The next row should happen inside tidy but it doesn't. Why? decisions <- decisions[ , colnames(decisions) != "Obs"] # BAU options are added to each combination Decisionmaker - Decision - Variable. This should be included in the DecisionTableParser. BAUs <- decisions[!duplicated(decisions[c("Decisionmaker", "Decision", "Variable")]), ] BAUs$Change <- "Add" BAUs$Result <- 0 BAUs$Option <- "BAU" decisions <- rbind(BAUs, decisions) DecisionTableParser(decisions) Fetch2(data.frame(Name = c("health.impact", "exposure"), Key = c("h796UrJF1UAVZp9H", "m22i7AfzxByOBaaG"))) # Evaluate ovariables and add decisions to them. This part should be put inside ComputeDependencies. exposure <- EvalOutput(exposure) exposure <- CheckDecisions(exposure) health.impact <- EvalOutput(health.impact) health.impact <- CheckDecisions(health.impact) summary(exposure) summary(health.impact) # Stakeholder probabilities are not implemented yet. ########### Graphs about health impacts by different decision options or source of estimates. ggplot(health.impact@output, aes(x = Year, y = health.impactResult, colour = Health.promotion)) + geom_boxplot() + theme_grey(base_size = 24) ggplot(health.impact@output, aes(x = Year, y = health.impactResult, colour = Cleaning.policy)) + geom_boxplot() + theme_grey(base_size = 24) ggplot(health.impact@output, aes(x = Year, y = health.impactResult, colour = health.impactSource)) + geom_boxplot() + theme_grey(base_size = 24) ######################### Calculate endpoints of interest for each stakeholder. endpoints <- tidy(opbase.data("Op_en5677.endpoints")) # List of stakeholders' endpoints. # Remove these redundant columns from intermediate results: removals <- c("exposureUnit", "exposureDescription", "exposureSource", "health.impactDescription", "health.impactSource", "health.impactUnit", "exposureResult", "health.impactResult") endpoint <- list() # Initiate a list for collecting stakeholder-specific endpoints. for(i in unique(endpoints$Stakeholder)) { print(i) dectable <- endpoints[endpoints$Stakeholder == i, ] # Initiate an endpoint ovariable for the next stakeholder. Results will come later. endpoint[[i]] <- new("ovariable", name = "endpoint", output = data.frame(Result = 0)) for (j in 1:nrow(dectable)) { # In the decision table format conditions are given in the "Cell"-column separated by ";". sel1 <- strsplit(as.character(dectable[j, "Cell"]), split = ";")[[1]] # ":" defines index - location matches as a condition. sel2 <- strsplit(sel1, split = ":") # No need for lapply, since strsplit is a vectorized function and current list depth is 1. # Create a list of conditions which the decision and option specific condition vector consists of. tempovar <- get(as.character(dectable[j, "Variable"])) # In this particular case, take only the results derived by Formula, because data is poor. temp <- tempovar@output[tempovar@output$health.impactSource == "Formula", ] for (k in 1:length(sel1)) { # For each condition separated by ";" if (length(sel2[[k]]) > 1) { # If ":" has been used for condition k locs <- strsplit(sel2[[k]][2], split = ",")[[1]] # Split by "," for multiple locs per given index temp <- temp[temp[, sel2[[k]][1]] %in% locs , ] # Match our data.frame to the condition temp <- temp[, colnames(temp) != sel2[[k]][1] ] # Remove all indices that were used in selecting rows, because otherwise they cannot be merged. } } # Make an ovariable out of the rows matching the condition. tempovar@output <- temp # Multiply by the weight and add to previous stakeholder endpoint. tempovar <- tempovar * as.numeric(as.character(dectable[j, "Result"])) # Remove columns that are not needed but may confuse merge. tempovar@output <- tempovar@output[ , !colnames(tempovar@output) %in% removals] endpoint[[i]] <- endpoint[[i]] + tempovar } } endpoint ############### Make graphs about the endpoint by decision options ggplot(endpoint[[1]]@output, aes(x = Cleaning.policy, y = Result, colour = Health.promotion)) + geom_boxplot() + theme_grey(base_size = 24) + labs( # label names title = names(endpoint)[1], y = "Euro", x = "Cleaning policy decision" ) ggplot(endpoint[[2]]@output, aes(x = Health.promotion, y = Result, colour = Cleaning.policy)) + geom_boxplot() + theme_grey(base_size = 24) + labs( # label names title = names(endpoint)[2], y = "Euro", x = "Health promotion decision" )

See also

• Descriptions of a previous structure
• ----#: . Päätöksenteon sokea piste: se mitä ihmiset eivät näe mutta eivät myöskään huomaa etteivät näe. Kuitenkin tutkimalla sitä mitä mitä ihmiset eivät näe saadaan selville asioita sokeasta pisteesta. Ymmärtämällä sokeaa pistettä voidaan keksiä asioita jotka järjestelmällisesti jäävät huomaamatta ja asioita, joilla voidaan korjata järjestelmällisiä puutteita. Avoin arviointi on tämmöinen päätöksenteon järjestelmällisten puutteiden korjausmekanismi. --Jouni 08:55, 1 May 2012 (EEST) (type: truth; paradigms: science: comment)

Related files

<mfanonymousfilelist></mfanonymousfilelist>