Timeline for Plotting summary statistics with mean, sd, min and max?
Current License: CC BY-SA 3.0
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 3, 2014 at 15:30 | comment | added | Ridhima | God knows how that happened, sorry about this :-) | |
| Feb 3, 2014 at 15:30 | vote | accept | Ridhima | ||
| Feb 2, 2014 at 15:19 | comment | added | TWL | @Ridhima, I have noticed that you have unchecked this answer after a while, could you let me know the reason? I'd greatly appreciate your feedback. | |
| Dec 3, 2013 at 15:39 | vote | accept | Ridhima | ||
| Feb 1, 2014 at 12:30 | |||||
| Dec 1, 2013 at 22:31 | comment | added | TWL | @MichaelMayer, many thanks for this example. It illustrates the difference nicely. | |
| Nov 28, 2013 at 15:11 | comment | added | Ridhima | Thanks all, I like the proposed violin plots so will go ahead with that choice :-) | |
| Nov 27, 2013 at 18:17 | history | edited | TWL | CC BY-SA 3.0 |
added 25 characters in body
|
| Nov 27, 2013 at 18:08 | history | edited | TWL | CC BY-SA 3.0 |
added 4 characters in body
|
| Nov 27, 2013 at 18:04 | comment | added | Michael M | @TWL: The topic is much too broad to discuss here. But take for instance economic evaluations of drugs: For the patient, maybe it's important to know median treatment duration, while for the health insurance company it's the mean treatment duration because they need to pay it for each patient. A curious fact: In the case of the exponential distribution, mean +/- 1 standard deviation covers 68% of all mass, mean +/- 2 sds covers about 95% of all mass. As for the normal. (But it's mere chance ;)) | |
| Nov 27, 2013 at 18:02 | history | edited | TWL | CC BY-SA 3.0 |
deleted 35 characters in body
|
| Nov 27, 2013 at 17:42 | comment | added | TWL | @MichaelMayer, regarding the summary.data.frame, I am not sure why sd is not reported. My concern about plotting mean +/- sd is that it indicates a symmetric distribution of values around mean. This is not true for some skewed distributions, in which case the position of the median with respect to the IQR would be more valuable. Hence, the summary statistic should be primarily chosen based on the underlying distribution. | |
| Nov 27, 2013 at 17:37 | comment | added | Nick Cox | Indeed, sometimes you need to see the mean and the SD so that you judge whether they are useful.... | |
| Nov 27, 2013 at 17:17 | history | edited | TWL | CC BY-SA 3.0 |
deleted 5 characters in body
|
| Nov 27, 2013 at 17:06 | history | edited | TWL | CC BY-SA 3.0 |
added 330 characters in body
|
| Nov 27, 2013 at 16:31 | comment | added | Michael M | Depending on the purpose of the analysis, mean and standard deviation is exactly what you need. I don't understand however the inconsistancy in R's summary.data.frame. It shows means but no sds. I can't think of many situations where means are useful but standard deviations misleading. | |
| Nov 27, 2013 at 15:48 | history | edited | TWL | CC BY-SA 3.0 |
added 968 characters in body
|
| Nov 27, 2013 at 15:42 | comment | added | Roland | I'd prefer a violin plot over this. | |
| Nov 27, 2013 at 15:28 | review | First posts | |||
| Nov 27, 2013 at 15:30 | |||||
| Nov 27, 2013 at 15:12 | history | answered | TWL | CC BY-SA 3.0 |