0
$\begingroup$

I have data with which I plot a histogram.

I would like to know the standard deviation of each bar of histogram.

Do you think it is an acceptable method to divide the samples into sub-samples for which I calculate the height of a bar (and so I will get a new sample of heights), and with this new sample I calculate a standard deviation?

OR should I use an alternative method I found (i.e., bootstrapping). In that case, many new samples are created from the original one, I could get the heights for each of them and then I could get stand deviation for each height.

  • What can be told about the accuracy of these standard deviations?
  • Are they OK?
  • Do you know a better method?
$\endgroup$
  • $\begingroup$ Are you referring to the st.dev of the data within each bar, or are you wanting the standard deviation of your measurement methods associated with that range of data? It is important to note that whatever you choose, the st.dev is really only used to describe normal data, so you should have a justification for why the data your are describing can be assumed to be normal. $\endgroup$ – Dinre Feb 6 '13 at 20:10
  • $\begingroup$ I would say that the SD is a perfectly fine measure of the variability in a dataset that can be used regardless of the distribution, @Dinre. I would just say that, (a) for many distributions (eg Poison), the SD is not an independent piece of information as it is a function of the mean, & (b) a SD could be misleading in some circumstances if people assume / interpret it as though it had come from a normal (eg, +/- 1 SD = 68% only for the normal), but this is a problem if people automatically assume that, it's not a problem intrinsic to the SD itself. $\endgroup$ – gung - Reinstate Monica Feb 6 '13 at 21:38
  • $\begingroup$ There is an answer but that just looks the same, it is not same question. The indent is different behind the question. $\endgroup$ – Aftershock Feb 7 '13 at 11:37
  • 1
    $\begingroup$ I want the st.dev of the data within each bar $\endgroup$ – Aftershock Feb 7 '13 at 11:39
  • $\begingroup$ Please help us here: if the apparent duplicate "looks the same," then--it's a duplicate! We cannot fathom your intent unless you expressly communicate it to us. Could you please point out why this is different and how any answers would need to be different than existing ones? $\endgroup$ – whuber Feb 8 '13 at 3:31