在R的simstudy包中实现条件分类分布时遇报错求助

I’ve run into this exact issue with simstudy before! The error "Categorical distribution requires 2 or more probabilities" pops up because when defining a conditional categorical variable, every condition you specify needs to provide at least two probabilities (since categorical variables must have 2+ categories).

Let’s break down what went wrong and fix your code. First, here’s a minimal example that reproduces your error (matching your initial code structure):

library(simstudy)

# Define non-random xNr
def <- defData(varname = "xNr", dist = "nonrandom", formula = 7, id = "idnum")

# This will throw the error: only one probability specified for the condition
def <- defData(def, varname = "contributor", dist = "categorical", formula = "0.8 | xNr ==7")

dat <- genData(100, def)

The problem here is that we’re only giving one probability (0.8) for the condition xNr ==7—but simstudy needs probabilities for all categories (at least two) to generate the categorical variable.

Fix 1: Use conditional probability vectors with ifelse #

If you want contributor to have two categories (e.g., "Yes" and "No") with conditional probabilities based on xNr, you can pass a vector of probabilities in the formula using ifelse:

library(simstudy)

def <- defData(varname = "xNr", dist = "nonrandom", formula =7, id = "idnum")

# Define contributor with two categories, conditional on xNr
def <- defData(def, varname = "contributor", dist = "categorical",
               formula = "c(0.7, 0.3)",  # Probabilities for "Yes" and "No" when xNr=7
               levels = c("Yes", "No"))

# Generate data
dat <- genData(100, def)

# Check results
table(dat$contributor)

If xNr had multiple values, you’d extend the ifelse to handle each case:

# Let's make xNr have two possible values first
def <- defData(varname = "xNr", dist = "categorical", formula = "0.5;0.5", levels = c("7", "9"))

# Conditional probabilities for each xNr value
def <- defData(def, varname = "contributor", dist = "categorical",
               formula = "ifelse(xNr == '7', c(0.7, 0.3), c(0.2, 0.8))",
               levels = c("Yes", "No"))

dat <- genData(100, def)
table(dat$xNr, dat$contributor)

Fix 2: Use defDataCond for explicit conditional definitions #

Another clean approach is to use defDataCond to define separate probability rules for each condition:

library(simstudy)

def <- defData(varname = "xNr", dist = "nonrandom", formula =7, id = "idnum")

# Define conditional rule for xNr=7
defCond <- defDataCond(
  def,
  varname = "contributor",
  dist = "categorical",
  formula = "0.7;0.3",  # Probabilities for "Yes" and "No"
  levels = c("Yes", "No"),
  condition = "xNr ==7"
)

dat <- genData(100, defCond)

This is especially useful if you have multiple complex conditions to handle—it keeps your code more readable.

The key takeaway: whenever you define a categorical variable (conditional or not), you must provide a set of probabilities for all categories (at least two) that sum to 1.

内容的提问来源于stack exchange，提问作者Monody