rtnorm_0_1_90ci produces weird distribution with upper bound = 0.99

[larscaspersen]

Running the rtnorm_0_1_90ci function with the upper range set to 0.99 leads to a completely different distribution than with upper bound set to 0.98 for example.

Here is an example: `tnorm_90 <- decisionSupport::rtnorm_0_1_90ci(n = nsim, lower = 0.53, upper = 0.90 ,correlationMatrix = 1) tnorm_95 <- decisionSupport::rtnorm_0_1_90ci(n = nsim, lower = 0.53, upper = 0.95 ,correlationMatrix = 1) tnorm_98 <- decisionSupport::rtnorm_0_1_90ci(n = nsim, lower = 0.53, upper = 0.98 ,correlationMatrix = 1) tnorm_99 <- decisionSupport::rtnorm_0_1_90ci(n = nsim, lower = 0.53, upper = 0.99 ,correlationMatrix = 1)

tnorm_df <- rbind(data.frame(dist = tnorm_90, type = 'upper limit = 0.90'), data.frame(dist = tnorm_95, type = 'upper limit = 0.95'), data.frame(dist = tnorm_98, type = 'upper limit = 0.98'), data.frame(dist = tnorm_99, type = 'upper limit = 0.99'))

line_df <- data.frame(upper = c(0.9, 0.95, 0.98, 0.99), lower = rep(0.53), middle = c(mean(c(0.53, 0.90)), mean(c(0.53, 0.95)), mean(c(0.53, 0.98)), mean(c(0.53, 0.99))), type = c('upper limit = 0.90', 'upper limit = 0.95', 'upper limit = 0.98', 'upper limit = 0.99'))

ggplot(tnorm_df, aes(x = dist)) + geom_histogram(bins = 150) + geom_vline(data = line_df, aes(xintercept = upper), col = 'red')+ geom_vline(data = line_df, aes(xintercept = lower), col = 'red')+ geom_vline(data = line_df, aes(xintercept = middle), col = 'red')+ facet_wrap(~type, scales = 'free_y', ncol = 1, nrow = 4) + xlim(0,1) + theme_bw()`

This produces the following distributions

The redlines indicate the lower bound, the mean and the upper bound. As you can see, the last example produces a completely different distribution. No warning appears. I suspect the same distribution appears when running mcSimulation with a variable which distribution is set to be tnorm_0_1 and upper range is 0.99.

I think it would be safe, if the function returns an error which sais that upper range with 0.99 is not valid