What does a period in my correlation coefficient table mean? How should I address it when writing up my results?

[Jessie Dodge]

Dear Statlist forum,

I am running a pairwise correlation coefficient (pwcorr function) and there is a "." between two variables (see output below). I have looked in prior STATLIST responses as to what this could mean (see this post https://www.statalist.org/forums/for...ead-of-p-value), and it seems like it is a sign that there could be multicolinarity. This would make sense to me as one variable is a response about experiencing discrimination since joining an organization (MinSt~E_48RR) and the other it could be highly correlated with is about experiencing discrimination in the past 30 days in the organization (MinSt~_48RR). Specifically, MinSt~_48RR is a subset of those that marked they had experienced discrimination since joining the organization (MinSt~E48RR). I am confused because if it was highly correlated why wouldn't it be 1 or close to the value of 1 (i.e. .98 or .88)? How should I address the "." when writing up the results?

With gratitude for all that you do!


. pwcorr FamMinStress_Ever MinStressLIFE_48RR MinStress30_48RR

| FamMin~m M~E_48RR M~0_48RR
-------------+---------------------------
FamMinStre~m | 1.0000
MinSt~E_48RR | 0.4332 1.0000
MinSt~0_48RR | -0.1033 . 1.0000

[Clyde Schechter]

I wonder if there was a skip pattern in your data so that people didn't answer one of these two items if the answer to the other was negative, or something like that. Anyway, what I'm getting at, regardless of how it might have happened, is the possibility that one of these two variables is always missing, so there are no pairs of these variables to calculate a correlation with. What do you get if you run -count if !missing(MinStressLife_48RR, MinStress30_48RR)-. I suspect you are getting zero.

If I am wrong, please post back showing example data using the -dataex- command. If you are running version 16 or a fully updated version 15.1 or 14.2, -dataex- is already part of your official Stata installation. If not, run -ssc install dataex- to get it. Either way, run -help dataex- to read the simple instructions for using it. -dataex- will save you time; it is easier and quicker than typing out tables. It includes complete information about aspects of the data that are often critical to answering your question but cannot be seen from tabular displays or screenshots. It also makes it possible for those who want to help you to create a faithful representation of your example to try out their code, which in turn makes it more likely that their answer will actually work in your data.

[Bruce Weaver]

I started working out the following example before I noticed that Clyde had responded. I'll post it anyway, in case it is helpful. (I added the -count- command at the end after reading Clyde's post.)

sysuse auto, clear

generate byte odd = mod(_n,2)
generate byte even = !odd
tabulate odd even

generate x1 = price
generate x2 = headroom
generate x3 = mpg

pwcorr x1 x2 x3, obs

* Set x2 = . for odd numbered observations
replace x2 = . if odd
* Set x3 = . for even numbered observations
replace x3 = . if even
pwcorr x1 x2 x3, obs

count if !missing(x2,x3)

Here is the output from the last two commands:

. pwcorr x1 x2 x3, obs

             |       x1       x2       x3
-------------+---------------------------
          x1 |   1.0000
             |       74
             |
          x2 |   0.1737   1.0000
             |       37       37
             |
          x3 |  -0.3607        .   1.0000
             |       37        0       37
             |

.
. count if !missing(x2,x3)
  0

[Jessie Dodge]

Hi Clyde,

Thank you so much for your quick reply! You are correct in that there was a skip pattern. Meaning that respondents were only asked the second question if they responded "yes" to the first. That was it!

Jessie