How can I change the scale of the y-axis in a Barchart

[Julie Iloul]

Hello,

I would be very grateful if someone could help me with the scale of the y-axis of my barchart (please find it attached to this post).

Since the average values of the y-axis are all above 800, I wanted to re-scale the range with a min. of 600 and max. of 1,100 (just so it looks better).

The command I used was:

graph bar (mean) lnexpy, over(country, sort(lnexpy) label(labsize(small))) /// 
blabel(bar, size(small)) ytitle(Average lnexpy) ///
ytitle(, size(small)) title(Average Export Sophistication)

Would be great if you could help,
Thank you and best regards,
Julie

[Andrew Musau]

First of all, your variable is in logs, so a bar chart does not make sense. Even if you specify that the minimum value be 600, you could also have specified some other value such as 100. Since we use bar charts to compare heights of bars, starting points will greatly influence perceptions of differences between heights. It is accepted that bar charts start at 0 (or at least, the reference point is 0), which leads to no ambiguity if this is the case. But this starting point cannot work for a logarithmic scale. Therefore, look at other graphs such as dot charts.

[Nick Cox]

I agree with @Andrew Musau's general notion that bar charts are dubious for data on logarithmic scales, but the typical value of the variable plotted is around 1000. So if what are shown are natural logarithms, then the original values were around exp(1000), which I find hard to believe. So the prefix ln I guess arises for other reasons.

Here is the graph shown properly as .png -- as we ask explicitly in FAQ Advice #12 -- not as a .gph attachment, which is more awkward.

With a range from 896 to 1003 on whatever scale it is (who can interpret the first decimal place?) I suggest instead a dot chart -- and horizontal alignment for better readability

* Example generated by -dataex-. To install: ssc install dataex
clear
input str15 country float lnexpy
"Brunei"       954
"Cambodia"     896
"Indonesia"    961
"Laos"         929
"Malaysia"     986
"Myanmar"      936
"Philippines"  966
"Singapore"   1003
"Thailand"     982
"Vietnam"      965
end

graph dot (asis) lnexpy, over(country, sort(lnexpy) label(labsize(small))) blabel(bar, size(small)) ytitle("") title(Average Export Sophistication) exclude0 ///
linetype(line) lines(lc(gs12) lw(vthin)) ysc(alt r(880 1020)) scheme(s1color)

.

Much hinges on personal taste, but I also suggest that the ytitle() in #1 is redundant given the much clearer title.

[Andrew Musau]

but the typical value of the variable plotted is around 1000. So if what are shown are natural logarithms, then the original values were around exp(1000), which I find hard to believe.

Good point Nick. The graph in #1 is not visible in your post, so I upload it here.

[Nick Cox]

I can see it....

[Andrew Musau]

Because it is saved somewhere on your PC. This happens sometimes with uploading graphs to the forum. Try your smart phone or another PC.

[Julie Iloul]

Dear both,

Thank you so much for your help!

Sorry for attaching it as a .gph attachment!

The original values of EXPY were actually around exp(1000). That was one of the main reasons why I've chosen to use the natural logarithm.

Thanks again!
Julie

[Nick Cox]

I can see it on my smart phone too.

[Andrew Musau]

That is odd. This is what I see.