Expert: Josh - 4/6/2009

Question
I'm trying to get error bars on my bar chart on Excel.

My graph shows the mean egg production on each day of life for a bunch of samples (X = day of life, Y = eggs produced on that day).

Since the raw data is already the means, do I use the mean of the means, so to speak, for the standard deviation?

This wouldn't really make sense since the bars on the graph have no relevence to each other (e.g. egg production on Day 1 is separate from that on Day 2).

If not, I'm not sure how you can get a different error bar for each bar of the graph when it's the same dataset.

Thanks.

Answer
Hi Katie,

I think the average should be computed for each value of X, over all samples that belong to that class (i.e., all samples with the same X value).

For illustration, suppose the (Days, Eggs) statistics are gathered from three farms, at 100 day and 200 day intervals such that

Farm A: (X=100,Y=20),(200,50),
Farm B: (X=100,Y=40),(200,55),
Farm C: (X=100,Y=30),(200,60).

In this example, our dataset contains only 6 samples in total. The idea here is to find out how many eggs are produced on average 100 days after hatching, and then 200 days after hatching.

The mean would be calculated like this:
For X=100,
mean(Y)=(20+40+30)/3=30,
variance(Y)=((-10)^2+(10)^2+0^2)/3=66.667
stdev(Y)=sqrt(var(Y))=8.165.  <= this figure is used to draw the error

For X=200,
mean(Y)=(50+55+60)/3 and so forth.