The code block below generates 60 plausible responses so the
remainder of this vignette runs without a network connection. Replace
responses with the output of
read_sheet_responses() and re-run from the quality section
onward to use real data.
set.seed(2024)
n <- 60
# Each construct is driven by a latent score plus item-level noise.
# Correlations between constructs are introduced by sharing variance.
lat <- function(mu, sigma) pmax(1, pmin(5, round(rnorm(n, mu, sigma))))
add_noise <- function(x, sigma = 0.45) {
pmax(1L, pmin(5L, as.integer(round(x + rnorm(n, 0, sigma)))))
}
# Higher service quality and sustainability lift satisfaction.
lat_dsqa <- lat(3.7, 0.6)
lat_dsqt <- lat(3.6, 0.6)
lat_dsuq <- lat(3.5, 0.6)
lat_ts <- pmax(1, pmin(5, round(
0.4 * lat_dsqa + 0.3 * lat_dsqt + 0.2 * lat_dsuq + rnorm(n, 0.6, 0.3)
)))
lat_bi <- pmax(1, pmin(5, round(0.7 * lat_ts + rnorm(n, 0.5, 0.4))))
# Digital marketing constructs are loosely correlated with each other.
lat_dmre <- lat(3.8, 0.6)
lat_dmau <- pmax(1, pmin(5, round(0.5 * lat_dmre + rnorm(n, 1.9, 0.4))))
lat_dmeu <- pmax(1, pmin(5, round(0.4 * lat_dmre + rnorm(n, 2.2, 0.4))))
lat_dmpv <- pmax(1, pmin(5, round(0.3 * lat_dmre + rnorm(n, 2.5, 0.4))))
# Repeat visitors score slightly higher on satisfaction and intention.
visitor_type <- sample(c("first_time", "repeat"), n,
replace = TRUE, prob = c(0.45, 0.55))
lat_ts[visitor_type == "repeat"] <- pmin(5L, lat_ts[visitor_type == "repeat"] + 1L)
lat_bi[visitor_type == "repeat"] <- pmin(5L, lat_bi[visitor_type == "repeat"] + 1L)
# Build item columns from latent scores.
make_cols <- function(lat, k, prefix) {
setNames(
as.data.frame(
vapply(seq_len(k), function(i) add_noise(lat), integer(n))
),
paste0(prefix, seq_len(k))
)
}
sim_df <- cbind(
data.frame(
respondent_id = sprintf("R%03d", seq_len(n)),
submitted_at = format(
seq(as.POSIXct("2025-01-10 09:00", tz = "UTC"),
by = "1 hour", length.out = n),
"%Y-%m-%dT%H:%M:%SZ"
),
started_at = format(
seq(as.POSIXct("2025-01-10 08:50", tz = "UTC"),
by = "1 hour", length.out = n),
"%Y-%m-%dT%H:%M:%SZ"
),
gender = sample(c("male", "female", "transgender"),
n, TRUE, c(0.44, 0.55, 0.01)),
age_band = sample(c("18_25", "26_40", "41_50", "51_60", "60_plus"),
n, TRUE, c(0.15, 0.40, 0.25, 0.15, 0.05)),
visitor_type = visitor_type,
freq_visit = sample(c("lt_1", "once", "2_3", "gt_3"),
n, TRUE, c(0.15, 0.25, 0.35, 0.25)),
nationality = sample(
c("thai", "chinese", "japanese", "korean", "indian",
"australian", "british", "german", "american", "other"),
n, TRUE),
education = sample(c("high_school", "diploma", "undergraduate", "post_graduate"),
n, TRUE, c(0.05, 0.10, 0.48, 0.37)),
profession = sample(c("student", "business", "salaried", "freelancer", "not_working"),
n, TRUE, c(0.15, 0.20, 0.45, 0.12, 0.08)),
companion = sample(c("friends", "family", "other"),
n, TRUE, c(0.35, 0.55, 0.10)),
group_size = sample(c("lt_3", "3_5", "gt_5"),
n, TRUE, c(0.30, 0.50, 0.20)),
stringsAsFactors = FALSE
),
make_cols(lat_dmre, 5, "dmre_"),
make_cols(lat_dmau, 5, "dmau_"),
make_cols(lat_dmeu, 5, "dmeu_"),
make_cols(lat_dmpv, 5, "dmpv_"),
make_cols(lat_dsqa, 4, "dsqa_"),
make_cols(lat_dsqt, 5, "dsqt_"),
make_cols(lat_dsuq, 11, "dsuq_"),
make_cols(lat_ts, 3, "ts_"),
make_cols(lat_bi, 3, "bi_")
)
# Align the data frame to the instrument.
responses <- read_responses(sim_df, study,
respondent_id = "respondent_id",
submitted_at = "submitted_at",
meta_cols = "started_at",
strict = FALSE)
cat("Respondents:", nrow(responses), "\n")
#> Respondents: 60
cat("Columns: ", ncol(responses), "\n")
#> Columns: 58
