RE: RE: st: SAS Stata equivalent command

["b. water" <barleywater@hotmail.com>]

dan, thank you for your advice. the website reference no doubt would be 
useful. although your suggestion did perhaps elucidate the particular 
portion of the SAS codes, overall it adds to my confusion. i am not sure how 
much of the paper i can copy in here without running foul of copyright to 
help me with the discussion but i am going out on a limb anyway:

before the juncture that i am stuck, i have created a bi_meta.dta data set 
based on the following command. the previous file 'meta' (set meta) was 
merely an infile of the data coupled with generate logit and variance of the 
logit of the sensitivity and specificity.

============================================

/*
In the next step we rebuild the original dataset so that each study will now 
have two records: one record holding (logit) sensitivity and the other 
(logit) specificity. Define two indicator variables: dis and non_dis. When a 
record holds logit sensitivity the value for dis=1 and for non_dis=0, when a 
record holds specificity dis=0 and non_dis=1
*/

data bi_meta;
set meta;

/* creating the record for logit sensitivity */
dis = 1; non_dis = 0;
logit = log_sens; var_logit = var_log_sens;
rec=1;
output;

/* variable rec uniquely identifies each record */
/* writes the record holding sensitivity to the new dataset */

/* creating the record for logit sprecificity */
dis = 0; non_dis = 1;
logit = log_spec; var_logit = var_log_spec;
rec=1; output;
run;

============================================

my understanding is this is akin to -reshape long- i.e. from having 
variables log_sens var_log_sens log_spec and var_log_spec, i have obtained a 
'longer' dataset with variables logit var_logit dis non_dis (where the 
combination of dis and non_dis would code for sensitivity and specificity) 
in preparation for the PROC Mixed.

it is following this step that your suggestion confused me. it does not make 
sense to simply concatenate three zero's onto var_logit in bi_meta.dta 
dataset and without SAS i cannot try to emulate the codes to visualise the 
SAS routines and adapt it for Stata.

following creation of the three zero's the codes called for SAS Proc Mixed 
module and proceeded as follows:

============================================

/*
Use the Proc Mixed module in SAS is used to set up the bivariate model. We 
apply the approach of Van Houwelingen et al to incorporate both the within 
and between study variance, more details and explanations can be found 
there.{van Houwelingen, 2002 #2} Use the restricted maximum likelihood 
estimation (REML) method in estimating the model
*/

proc mixed data=bi_meta method=reml cl ; /* option cl will give confidence 
intervals */
/* study_id and modality are categorical variables */
class study_id modality;

/* model statement: asking for different estimates of mean sensitivity and 
specificity for each  modality, provide large value for degrees of freedom 
to obtain p-values based on normal distribution rather than the 
t-distribution (=default in SAS) */

model logit = dis*modality non_dis*modality / noint cl df=1000, 1000, 1000, 
1000, 1000, 1000;

/* random effects for logit sensitivity and specificity with possible 
correlation (UN=unstructured covariance structure) */
random dis non_dis / subject=study_id type=un ;

/* use the repeat statement to define different within-study variances for 
sens and spec in each study */
repeated / group=rec;

/* name the file holding the all the (co)variances parameters, keep the 
within-study variance constant */
parms / parmsdata=cov hold=4 to 91;

/* use contrast statement for testing specific hypotheses */
/* testing for differences in sensitivities */
contrast �CT_sens vs LAG_sens� dis*modality 1 -1 0 / df=1000 ;
contrast �CT_sens vs MRI_sens� dis*modality 1 0 -1/ df=1000 ;
contrast �LAG_sens vs MRI_sens� dis*modality 0 1 -1/ df=1000 ;

/* testing for differences in specificities */
contrast �CT_spec vs LAG_spec� non_dis*modality 1 -1 0 / df=1000 ;
contrast �CT_spec vs MRI_spec� non_dis*modality 1 0 -1/ df=1000 ;
contrast �LAG_spec vs MRI_spec� non_dis*modality 0 1 -1/ df=1000 ;

/* testing for differences in DOR */
contrast �CT_odds vs LAG_odds � dis*modality 1 -1 0 non_dis* modality 1 -1 0 
/ df=1000 ;
contrast �CT_odds vs MRI_odds � dis* modality 1 0 -1 non_dis* modality 1 0 
-1 / df=1000 ;
contrast �LAG_odds vs MRI_odds � dis* modality 0 1 -1 non_dis* modality 0 1 
-1 / df=1000 ;
run;

============================================

as nick (cox) said in earlier posting, PROC Mixed equivalent could be either 
gllamm or xtmixed (Stata 9), but without being able to prepare the data for 
gllamm (since my Stata is version 8.2) i think i am effectively stuck for 
the time being.

once again though, thank you for your input dan, and the pointer to the 
website.

kind regards,
bw



> From: Dan Blanchette <dan_blanchette@unc.edu>
> Reply-To: statalist@hsphsun2.harvard.edu
> To: statalist@hsphsun2.harvard.edu
> Subject: RE: RE: st: SAS Stata equivalent command
> Date: Thu, 8 Dec 2005 13:00:54 -0500 (EST)
>
> Please check out "A SAS User's Guide to Stata" 
> www.cpc.unc.edu/services/computer/presentations/sas_to_stata
> where you can find some SAS code matched to the equivalent Stata code.
>
> In your circumstance, you have fallen into some obscure SAS coding.
> The code: 
> *************************************************************************
> if _n_ eq 1 then do;
>  est = 0; output; est = 0; output; est = 0; output;
> end;
> *************************************************************************
>
> creates a variable called "est" and sets it to 0 and then creates 3 
> observations
> in the dataset.  Why this is important is beyond me, but perhaps it is.
> Next the code brings in an existing data set "bi_meta" and concatenates 
> that
> onto these 3 observations.  Then it sets the variable "est" to be the same 
> value
> as the variable "var_logit."  And finally it only keeps this new variable 
> "est."
>
> So the eqivalent Stata code would be:
>
> *************************************************************************
>
> clear
> set obs 3
> gen est = 0
> append using "bi_meta.dta"
> replace est = var_logit
> keep est
>
> *************************************************************************
>
> I hope this helps,
>
> Dan Blanchette
>
> ITS Research Computing Group
> University of North Carolina Chapel Hill
> Dan_Blanchette@unc.edu
_________________________________________________________________
Don�t just search. Find. Check out the new MSN Search! 
http://search.msn.click-url.com/go/onm00200636ave/direct/01/

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/