About bncoxuk

Reeza · ‎08-09-2011

You never specified what your sample sizes are for each group. If your dealing with smaller numbers then having a large group (group2+3) vs group 1 will always give you a better fit because you're using more data. Just something to keep in mind.

StatDave · ‎08-05-2011

Maximum likelihood estimation does not involve sums of squares.

trekvana · ‎08-09-2011

bncoxuk- proc logit can run multinomial models. use link=glogit in the model statement options

Reeza · ‎08-02-2011

Have you tried looking at the Cheese example in the SAS Doc? Example 51.3 in SAS 9.2 documentation, similar situation or are you unsure how to specifically interpret your results?

Ksharp · ‎08-02-2011

data test; do index=1 to 100; y=int(ranuni(-1)*2); output; end; run; proc sort data=test ; by descending index; run; %let dsid=%sysfunc(open(test)); %let nobs=%sysfunc(attrn(&dsid,nobs)); %let dsid=%sysfunc(close(&dsid)); %let want_percent=.2; %let want_top=%sysevalf(&nobs*&want_percent,integer); %put _user_; proc means data=test(obs=&want_top) nway noprint; var y; output out=want mean= sum=/autoname; run; Ksharp

data_null__ · ‎07-14-2011

Does it mean that if I really prefer the GLM matrix, there is no way that I can choose my preferred reference group? The group that comes last when the values of the CLASS variables are ordered, depending on the value of ORDER= class statement option, will be the reference group when PARAM=GLM. You can recode or reFORMAT the value to achieve the order you desire.

FloydNevseta · ‎07-13-2011

data_null_ gave you the answer to your question. If it's too confusing you could write the %else with %do...%end as you did for the %if statement, such as: %else %do: SET work.model work.data%sysfunc(putn(&i+1.,z2.)); %end; That eliminates the consecutive semicolons and your confusion. bncoxuk wrote: Hi data_null_, Can I kow why there need to be two separate semicolons for %ELSE and SET, rather than a single semicolon for a statement.

Ksharp · ‎07-14-2011

You also can split the origin dataset into two datasets ,one contains ind=1 and the other contains ind=0, then %do i=1 %to 4; proc surveyselect data=list_stock method=srs n=4 out=sample&i noprint; run; data sample&i; set sample&i ind_1; run; %end; Ksharp

Doc_Duke · ‎07-11-2011

That's one of the reasons that we do Bootstraps...to see which variables are consistently in or out of a model. As to whether you need to include occupation in the end, that is more subtle than just looking at statistical significance. Maybe it is needed for "face validity," etc. One of the things that you may want to look at is how is Occupation coded. If it is a class variable with lots of levels, then you may just have too many level with low frequencies and lumping the codes would provide you with more information and a more stable model. It could also be that you have an interaction term that you have not accounted for (e.g. is education also in the model?). Doc Muhlbaier Duke

art297 · ‎07-11-2011

The putn function simply enables you to specify a numeric format at run time. Since you indicated that the file names had leading zeros, that is why I suggested the z format, as it displays numbers with leading zeros.

bncoxuk · ‎07-11-2011

Hi everyone, Thanks for help. I tried every answer and all worked. So all are correct answer. The forum system cannot give me the rating of correct answer for all responses. So I just randomly pick one and make the others as helpful.

DLing · ‎07-05-2011

You are right that as sample size increases, the ability to resolve increases, and you are able to split hair if you have huge data volume. Regression with multi-million records are being done routinely, something that's just not imaginable not that long ago. At some point one need to stop asking only statistical significance (is it really there or not?) and start asking context significance (so that 0.00001% difference is truly there, so what?). Ruth's point #4 last sentence is absolutely bang-on.

chang_y_chung_hotmail_com · ‎06-24-2011

in case bncoxuk wanted to "recode" using a format: /* a test dataset */ data one; a=18; b=19; output; a=15; b=24; output; run; /* recoding using a format */ proc format; value level 0-< 5 = 1 5-<10 = 2 10-<15 = 3 15-<20 = 4 20-<25 = 5 other = .; run; /* recoding */ data two; set one; a = input(put(a, level.), best.); b = input(put(b, level.), best.); run; /* check */ proc print data=two; run; /* on lst Obs a b 1 4 4 2 4 5 */

sbb · ‎06-14-2011

I have to ask why you double-posted your query/problem in this forum? Recommend you stay with one post, one topic-point or issue/problem/query....suggested forum etiquette. Here's your first one: http://support.sas.com/forums/thread.jspa?threadID=14647 Scott Barry SBBWorks, Inc. Message was edited by: sbb

SPR · ‎06-15-2011

Hello Bncoxuk, Check your LOG. It looks like you have an error: ELSE THEN should be simply ELSE. Sincerely, SPR

Online Status	Offline
Date Last Visited	‎09-01-2015 07:11 AM

Re: How to concatenate character values across variables

How to concatenate character values across variables

How to concatenate character values across variables

How to create a report by combining results from simple PROC SUMMARY

LIKE '%' does not work

What is the functionality of || here

What is the functionality of || here

What is the functionality of || here

Re: How to get the percentage for each unique combinations of variable...

How to get the percentage for each unique combinations of variables

How to set all missing values to zero for all variables?

Nominal logit models vs ordinal logit models

PROC CATMOD: type I or type III?

Does SAS has procedure with automatic variable selection method for mu...

The intercepts of a cumulative logit model

Re: How to get a summary for the top 5%, 10% and other data

Logistic procedure: reference group

Why two consecutive semicolons are needed?

How to draw n random samples from the original dataset

A variable not consistently significant over tests

How can I use a single program to run multiple data sets?

How to summarize the frequency distribution for all variables

Re: Sample size n and significance

Re: Is there any more clever way to do it?

Re: Why no output for cases with age<=3 ?

Re: Why failed to output?