topic Re: I am trying to obtain the crude ORs for several variables. in SAS Programming

I am trying to obtain the crude ORs for several variables.

deme_bb — Mon, 22 Jul 2019 14:49:33 GMT

I am trying to obtain crude OR and their relevant CIs for several variable using the procfreq tables function; however, the statement is not giving me odds ratios.

Proc Freq data=df_epi ;
tables age * Stroke / cmh;
run;

I obtained the adjusted OR using this code.

Proc Logistic data=df_epi;
class educ ;
model Stroke (Event='1') = educ Age bmi diabetes prevchd prevhyp sysbp diabp glucose totchol heartrte;
run;

Help please

Re: I am trying to obtain the crude ORs for several variables.

Reeza — Mon, 22 Jul 2019 15:47:12 GMT

From the documentation:

CMH	Requests all Cochran-Mantel-Haenszel statistics
CMH1	Requests CMH correlation statistic, adjusted odds ratios,
	and adjusted relative risks
CMH2	Requests CMH correlation and row mean scores (ANOVA)
	statistics, adjusted odds ratios, and adjusted relative risks

OR	Requests the odds ratio for tables

You haven't used an option that would provide OR, change your code by adding one of the following options instead and see if you get what you need.

https://documentation.sas.com/?docsetId=procstat&docsetTarget=procstat_freq_syntax08.htm&docsetVersion=9.4&locale=de

If your table is not a 2x2 table, this gets a bit trickier and you're likely better off just using PROC LOGISTIC to get the unadjusted rates.

Not sure how familiar you are with PROC LOGISTIC and logistic regression, but you may also want to specify PARAM=REF on the CLASS statement to get referential parameterization of your dummy variables, the default is GLM.

@deme_bb wrote:

I am trying to obtain crude OR and their relevant CIs for several variable using the procfreq tables function; however, the statement is not giving me odds ratios.

Proc Freq data=df_epi ;
tables age * Stroke / cmh;
run;

I obtained the adjusted OR using this code.

Proc Logistic data=df_epi;
class educ ;
model Stroke (Event='1') = educ Age bmi diabetes prevchd prevhyp sysbp diabp glucose totchol heartrte;
run;

Help please

Re: I am trying to obtain the crude ORs for several variables.

deme_bb — Mon, 22 Jul 2019 15:49:29 GMT

Hi Reeza,

My table is not a 2x2. Proc logistic would only give me the adjusted ratios correct? If it doe sindeed give me crude ratios im not sure on how to go about obtaining them.

Re: I am trying to obtain the crude ORs for several variables.

Reeza — Mon, 22 Jul 2019 15:53:15 GMT

Crude ratio's mean you haven't adjusted for any other variables. So if you run a logistic regression with a single variable that would be the crude rate.

Re: I am trying to obtain the crude ORs for several variables.

deme_bb — Mon, 22 Jul 2019 16:01:48 GMT

Okay so maybe i have it backwards. How would i determine my adjusted ratios?

The code below is the code i used to obtain the unadjusted ratios:

Proc Logistic data=df_epi;
class educ ;
model Stroke (Event='1') = educ Age bmi diabetes prevchd prevhyp sysbp diabp glucose totchol heartrte;
run;

So how would i obtain the adjusted ratios?

Re: I am trying to obtain the crude ORs for several variables.

Reeza — Mon, 22 Jul 2019 16:05:11 GMT

That is the adjusted ratios. Minus the fact you haven't fixed your class, unless you want that.

Re: I am trying to obtain the crude ORs for several variables.

deme_bb — Mon, 22 Jul 2019 16:07:06 GMT

What do you mean by "fix the class statement"?

SO my unadjusted odds ratio would be me running that statement with each individual variable separately?

Re: I am trying to obtain the crude ORs for several variables.

Reeza — Mon, 22 Jul 2019 16:11:08 GMT

@deme_bb wrote:

What do you mean by "fix the class statement"?

Not sure how familiar you are with PROC LOGISTIC and logistic regression, but you may also want to specify PARAM=REF on the CLASS statement to get referential parameterization of your dummy variables, the default is GLM.

SO my unadjusted odds ratio would be me running that statement with each individual variable separately?

. So if you run a logistic regression with a single variable that would be the crude rate.

Re: I am trying to obtain the crude ORs for several variables.

deme_bb — Mon, 22 Jul 2019 16:41:21 GMT

Thank you so much!