For the third plot, in addition to ggstatsplot, you can directly create a heatmap using geom_tile.
I don't have experience with this sort of multivariate sensitivity analysis, but you can create the second graph with the ggraph package. Below is an example with fake data. If you can put your sensitivity data into a form like the graph object created below, then you can plot it with ggraph. I haven't used ggraph much, so I'm not sure if I'm even referencing the attributes correctly in creating the graph, but hopefully this will help get you started.
For the first plot, it looks like you could create it with geom_bar, geom_errorbar, and coord_polar, but it would help if you provided some sample data to plot.
library(igraph)
library(ggraph)
# Create a graph of ten nodes with all nodes connected
g = make_graph(c(combn(LETTERS[1:10], 2)))
# Set edge weights and node sizes and add them to the graph as attributes
set.seed(2)
edge.wt = runif(length(E(g)), 0,1)
node.size = runif(length(V(g)), 1,4)
g = g %>%
set_edge_attr("weight", value=ifelse(edge.wt > 0.5, edge.wt^4, 0)) %>%
set_vertex_attr("size", value=node.size)
ggraph(g, layout = 'linear', circular = TRUE) +
geom_edge_arc(aes(edge_width=E(g)$weight), colour=hcl(120,100,60)) +
geom_node_point(aes(size=V(g)$size), colour=hcl(120,100,40)) +
geom_node_text(aes(label=V(g)$name), colour="white") +
coord_fixed() +
theme_void() +
scale_edge_width_continuous(range=c(0,2)) +
scale_size_continuous(range=c(2,8), limits=c(0,5), breaks=1:4) +
guides(size=FALSE, edge_width=FALSE)

UPDATE: Here's an attempt at the first plot, using one of the data sets @RuReady linked to:
library(tidyverse)
d = read_delim("analysis_heavy_aliphatic-C-O.txt", n_max=410, delim=" ")
# Convert data to long format
dl = d %>% select(Parameter, S1, ST) %>%
gather(key, value, S1, ST) %>%
left_join(
d %>% select(Parameter, S1=S1_conf, ST=ST_conf) %>%
gather(key, conf, S1, ST)
)
# Set up data for plotting
pdat = dl %>% group_by(Parameter) %>%
filter(any(value>0.05)) %>%
ungroup %>%
mutate(key = fct_relevel(key, "S1", after=Inf)) %>%
arrange(key)
ggplot(pdat, aes(Parameter)) +
geom_hline(yintercept=xgrid, colour="grey80", size=0.3) +
geom_segment(data=data.frame(x=unique(pdat$Parameter), y=0, yend=0.33),
aes(x=x, xend=x, y, yend=yend), colour='grey30', size=0.3,
position=position_nudge(x=0.5)) +
geom_segment(data=data.frame(x=0.5), x=0.5, xend=0.5, y=0, yend=.33, colour='grey30', size=0.3) +
geom_col(aes(y=value, fill=key), width=0.5, position="identity") +
geom_errorbar(aes(ymin=value - conf, ymax=value + conf, colour=key),
position=position_dodge(0.2), width=0.2) +
geom_text(aes(label=Parameter, y=0.36), size=3, colour="grey30") +
annotate(y=xgrid, x=0, label=xgrid, geom="text", size=3) +
coord_polar(start=-2*pi/17 * 0.5) +
scale_y_continuous(limits=c(-0.2, 0.36), expand=c(0,0)) +
expand_limits(y=-0.1, x=-0.5) +
theme_bw() +
theme(panel.grid.major=element_blank(),
axis.text=element_blank(),
legend.title=element_blank(),
panel.border=element_blank(),
axis.ticks=element_blank(),
axis.title=element_blank(),
legend.position=c(0.5,0.5),
plot.background=element_rect(colour=NA)) +
labs(fill="", colour="") +
scale_fill_manual(values=hcl(c(15,195), 100, 70)) +
scale_colour_manual(values=hcl(c(15,195), 100, 40))
This might look better with dodged bars instead of overlapped bars. I've included only the altered lines of code below. The rest is the same.
geom_col(aes(y=value, fill=key), width=0.8, position=position_dodge(0.8)) +
geom_errorbar(aes(ymin=value - conf, ymax=value + conf, colour=key),
position=position_dodge(0.8), width=0.2) +