[Tom Salvitti]

thanks a lot

[Tom Salvitti]

Carlo Lazzaro , I want to ask you a question.
I compare the other groups of patients in the same way.
If p is <0.005 we can say that there is difference in -vocalization. and so i can say that there is a relationship (correlation) between groups (for instance childrenHR- and -adults) ? or what?

Thanks a million

[Tom Salvitti]

Is it correct to say that there is a correlation if p is <0.005?
Thanks a lot Carlo Lazzaro

[Carlo Lazzaro]

Tom:
you can say that there's evidence of a difference in vocalization between the two groups.

[Tom Salvitti]

and in term of correlation? relationship we can say anything?
sorry for my dubt...
thanks a million Carlo Lazzaro

[Tom Salvitti]

i ask this Carlo Lazzaro because iI 've to answer what kind of relationship exist between the groups??

thanks a lot

[Carlo Lazzaro]

Tom:

tabulate patients , gen(patients)
. corr vocalization patients1 patients1 patients2 patients3 patients4 patients5
(obs=69)

             | vocali~n patien~1 patien~1 patien~2 patien~3 patien~4 patien~5
-------------+---------------------------------------------------------------
vocalization |   1.0000
   patients1 |  -0.0945   1.0000
   patients1 |  -0.0945   1.0000   1.0000
   patients2 |  -0.0223  -0.0626  -0.0626   1.0000
   patients3 |   0.0859  -0.1515  -0.1515  -0.3176   1.0000
   patients4 |  -0.0153  -0.1026  -0.1026  -0.2151  -0.5211   1.0000
   patients5 |  -0.0351  -0.0752  -0.0752  -0.1577  -0.3820  -0.2587   1.0000


.

[Tom Salvitti]

Dear Carlo Lazzaro ,
Thank you so much Carlo for the help.
How to read the output you sent me ?? Thank you!!!

[Carlo Lazzaro]

Tom:
to investigate correlation and its statistical significance comparing all the variables each other, you can exploit your original idea -pwcorr- adding -bonferroni- correction plus the significance level of correlation coefficients to be starred.
Elaborating on my previous example:

. pwcorr vocalization patients1 patients2 patients3 patients4 patients5 , star(.05) bonferroni

             | vocali~n patien~1 patien~2 patien~3 patien~4 patien~5
-------------+------------------------------------------------------
vocalization |   1.0000
   patients1 |  -0.0945   1.0000
   patients2 |  -0.0223  -0.0626   1.0000
   patients3 |   0.0859  -0.1515  -0.3176   1.0000
   patients4 |  -0.0153  -0.1026  -0.2151  -0.5211*  1.0000
   patients5 |  -0.0351  -0.0752  -0.1577  -0.3820* -0.2587   1.0000

The correlation pattern is steadily negative and it reaches statistical significance when starred.
That said, I would still go -regress-.

[Tom Salvitti]

Dear Carlo thanks a lot.
My idea is to do "regress" (the example that i wrote you yesterday children and adult) and then pwcorr.
Is it correct my idea Carlo Lazzaro ?

[Carlo Lazzaro]

Tom:
-regress- is enough here, as it allows you to investigate difference between groups of patients (via -test-)-

[Tom Salvitti]

In my example I have: | vocali~n patien~1 patien~1 patien~2 patien~3 patien~4 patien~5 -------------+--------------------------------------------------------------- vocalization | 1.0000 patients1 | -0.3637 1.0000 patients1 | -0.0509 -0.1816 1.0000 patients2 | -0.3770* -0.1627 -0.2067 1.0000 patients3 | 0.5384* -0.1816 -0.2308 -0.2067 1.0000 patients4 | 0.0930 -0.1627 -0.2067 -0.1852 -0.2067 1.0000 patients5 | 0.0849 -0.1816 -0.2308 -0.2067 -0.2308 -0.2067 1.0000
It's strange for me the significative results isn'it?
It's significative the correlation between patients2 and vocalization? How can explain this?

Thanks a lot Carlo Lazzaro in advance


Best Tommaso

[Tom Salvitti]

pwcorr vocalizzazioni var71 var72 var73 var74 var75 var76 , star(.05) bonferroni

             | vocali~i    var71    var72    var73    var74    var75    var76
-------------+---------------------------------------------------------------
vocalizzaz~i |   1.0000 
       var71 |  -0.3637   1.0000 
       var72 |  -0.0509  -0.1816   1.0000 
       var73 |  -0.3770* -0.1627  -0.2067   1.0000 
       var74 |   0.5384* -0.1816  -0.2308  -0.2067   1.0000 
       var75 |   0.0930  -0.1627  -0.2067  -0.1852  -0.2067   1.0000 
       var76 |   0.0849  -0.1816  -0.2308  -0.2067  -0.2308  -0.2067   1.0000

Sorry for the presentation of data..
I ve this situation...It's strange isn'it?..
How can explain this?
Thanks a million Carlo Lazzaro

[Carlo Lazzaro]

Tom:
it may be due to the mean and the standard deviation of the two variables.
Again, correlation is unuseful for what you've in mind.
Go -regress-.

[Tom Salvitti]

thank you Carlo Lazzaro