[Jdoggrr]

When placing error bars on a column graph with a logarithmic Y axis, the
error bar is drawn as a large dark triangle connecting the last bar and it's
neighbor. There is an error, something about negative and zero values in
logarithmic scales, but there are no such values in the data. Is there a
work-around for this?

[ShaneDevenshire]

Hi,

Send me a copy of the chart and I will look at it when I get home tonight.
--
Thanks,
Shane Devenshire
Microsoft Excel MVP


"Jdoggrr" wrote:

When placing error bars on a column graph with a logarithmic Y axis, the
error bar is drawn as a large dark triangle connecting the last bar and it's
neighbor. There is an error, something about negative and zero values in
logarithmic scales, but there are no such values in the data. Is there a
work-around for this?

[[email protected]]

Check your Min and Max values to make sure the DIFFERENCE between the
plotted value and +/- error value is not zero (which can't be plotted
on a log axis). In other words, the max or min error value cannot be
the same as the plotted value for any data point.

All the Best!

[callicles]

I think I just figured this out. I was trying to create semilog scatter
plots (log scale on y axis) with custom error bars. My y-axis range was
typically between 1 and 0.001. I was getting the same black triangle problem
on both new Excel 2007 plots and even on old plots from 2003 that I opened in
2007. What I just tried was to add the error bars BEFORE applying the log
scale on the y axis. IT worked! The error bars are scaled appropriately so
that equal magnitude bars are drawn with positive bars shorter than negative
ones. I also saved my test plot in Excel and reopened it. The correctly
drawn error bars were preserved.

I am able to copy these graphs to the clipboard, paste them into Word, save
and reopen the Word file, and even edit the data from within Word. Also, If
I change the magnitude of the error bar to greater than or equal the
magnitude of the data point, I get a warning about plotting negative values
on a log scale and a black triangle.

I suspect that the original problem is an order of operations issue, so that
Excel is checking to see if your proposed error bar length is too large
BEFORE it adjusts the magnitude to match the log scale.

I cannot believe it took me this long to figure this out, but I have been
struggling with this for over a week while trying to finalize my dissertation.