R - ggplot2 geom_tile creates way different graph to heatmap - r

Question

I would like to draw a heatmap for some data set.
Originally, I used the heatmap functions with satisfiying results. Yet, as all my other graphs are done with ggplot2, I tried to do it with ggplot2 as well. Yet, I get weird graphs and I am not sure what the reason is.
I am given a dataset as follows:
> dput(B)
structure(list(`2001` = c(510, 15, 14, 9, 8, 11, 7, 5, -1, -3),
`2002` = c(397, -13, 5, 6, 12, -1, 0, 2, 5, 3),
`2003` = c(323, -6, -2, 1, 0, 6, -5, -2, 1, 4),
`2004` = c(133, -2, 2, -4, 0, 5, 8, -2, 0, 1),
`2005` = c(-100, -8, -6, -2, 1, 2, 2, 3, -6, -5),
`2006` = c(-114, -7, 2, -4, -2, 0, 1, 2, 4, -3),
`2007` = c(-130,-13, 0, 4, -3, -2, -1, 1, 2, 4),
`2008` = c(-38, -10, 4, 0, 3, 4, 2, 0, 0, 1),
`2009` = c(-194, -13, -5, -4, -3, -1, 0, 1, 1, 1),
`2010` = c(-202, -6, 0, -1, -5, -2, -3, -1, 2, -2)),
row.names = 0:9, class = "data.frame")
Now using,
> heatmap(as.matrix(B), Colv = NA, Rowv = NA, col = cm.colors(10))
I obtain the following heatmap, which looks reasonable.
On the other hand, with
C <- B
C$size <- row.names(C)
C <- melt(C, variable.name="year",id=c("size"))
ggplot(data = C) +
geom_tile(aes(x = year, y = size, fill = value)) +
scale_fill_gradientn(colors=cm.colors(10))
I get the following
Why is this? And how can I correct this such that I get a plot similar to the first one?

Answer 1

Scaling is applied in heatmap. So you need to scale your data (according to size, in this instance) to reproduce something similar, such as:
C2 <- C %>% group_by(size) %>% mutate(rescale = scale(value))
ggplot(data = C2) +
geom_tile(aes(x = year, y = size, fill = rescale)) +
scale_fill_gradientn(colors=cm.colors(10))
There are other options, depending on what you are trying to visualise. For example, have a look at limits and values in scale_fill_gradientn.

Answer 2

The issue is that size = 0 has much larger numbers than the other entries. So the second heat map is correct because the rest of the values are very close to zero in respect to the given scale. I would say that the second one is actually correct because you can see that outside of size = 0 the numbers are roughly the same, especially in respect to the large numbers found in that first row. I'm not quite sure how heatmap works, and it doesn't supply a scale, but I think the second graph make sense give the low variability outside of the first row.