Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: basis for selection of observations for nonparametric correlation?

From   Nick Cox <>
Subject   Re: st: basis for selection of observations for nonparametric correlation?
Date   Mon, 4 Jul 2011 08:47:30 +0100

Stata ignores observations in which _any_ value is missing. Zeros are
values like any other so far as Spearman correlation is concerned; in
any case they are ranked before the correlation is calculated, so the
fact that any values are zeros is hidden as well as immaterial.


On Mon, Jul 4, 2011 at 8:19 AM, Edith Marshall <> wrote:

> Please forgive my ignorance and, thus, need to post - I am new to
> STATA and rusty with my stats!
> I am running Spearman's rank correlation (nonparametric) on a data set
> of 167 observations.  I have some missing values and legitimate zeros.
>  After selecting the variables of interest and specifying that I want
> to see the the correlation coefficient, the number of observations,
> and the significance value, I get my correlation results table.  The
> number of observations is 27 for all variable comparisons.  When I
> check the box that forces all observations to be considered and re-run
> the analysis, I get between 24 and 167 observations used for the
> different comparisons.
> Can anyone tell me the basis for which observations are used (the
> default) when I don't force all of them in?  I need to decide which
> correlation results to use and don't understand how STATA has selected
> the 27 from the 167.  Obviously, I would prefer to have more
> observations but also want the results to be legitimate.

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index