topic Re: macro to declare variable and beta for regression in Statistical Procedures

macro to declare variable and beta for regression

eemrun — Fri, 25 Aug 2017 03:42:35 GMT

Rookie here and struggling to create a declaration to run regression for multiple independent variables. The dataset looks like the following

A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	V0	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	T1	T2	T3	T4	T5	T6	T7	T8	T9	T10	T11	T12	T13	T14	T15	T16	T17	T18	T19
1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0

As part of Proc MCMC, I want to specify the following a model in the following manner:

y = a + a_b1*A1 + a_b2 *A2 + a_b3*A3... + b_b1*V1 + b_b2*V2 + ... and so on

Is there a shortcut way of doing this for a large sets of A, V and T?

Re: macro to declare variable and beta for regression

Reeza — Fri, 25 Aug 2017 15:23:17 GMT

Is there a pattern here? If so, then you could likely write a macro loop.

You can also look at the | and @ usage in writing a MODEL statement.

Re: macro to declare variable and beta for regression

Rick_SAS — Fri, 25 Aug 2017 15:42:40 GMT

You don't need a macro. You can compute the mean value (eta) in a loop, as shown on p 4-5 of High & ElRayes (2017):

Re: macro to declare variable and beta for regression

eemrun — Sun, 27 Aug 2017 21:19:42 GMT

thanks. that paper is helpful. however i am still getting an error (ideally stupidity from my end!). I am using the following argument:

proc mcmc DATA=input4 diag=all dic
	propcov=quanew 
	nbi=5000 
	ntu=5000 
	nmc=10000 
	thin=20 
	plots(smooth)=all seed=19680409
	outpost=pdres;


	array od[46] 	a1-a11 v0-v11 t1-t23;
	array bd[46] 	ab_1-ab_11 vb_0-vb_11 tb_1-tb_23;
	

	parms (ab_1-ab_11 vb_0-vb_11 tb_1-tb_23) 0;
	parms sigma2 1;

	/* Prior distribution assumptions */
	prior sigma2 ~ igamma(shape = 100, scale= 76);
	prior ab_1-ab_11 vb_0-vb_11 tb_1-tb_23 ~ normal(mean=0, var=1e6);

	
	do i = 1 to 46;
		p = od[i] * bd[i];
	end;
	model d ~ normal(p, var=sigma2);
run;

The error says: All observations in the input data set contain missing values.

Any idea where I am going wrong?

Re: macro to declare variable and beta for regression

eemrun — Sun, 27 Aug 2017 21:36:52 GMT

hi. The pattern is: 3 variable types: A, V, T. A ranges from 1-11, V ranges from 0-11 and T ranges from 1-23 (e.g. A1, A2, V0, V1, T1, T2, etc.). The values for these variables are 0,1.

Re: macro to declare variable and beta for regression

Tom — Mon, 28 Aug 2017 00:25:13 GMT

Any idea where I am going wrong?

The error says: All observations in the input data set contain missing values.

The error means that for every observations in your input data there is at least one variable that has a missing value.

In the example data you posted the variable V11 was missing on all observations.

Re: macro to declare variable and beta for regression

ballardw — Mon, 28 Aug 2017 16:14:11 GMT

Unless specifically instructed to allow missing values all the regression procedures default to dropping any record with any of the variables on the model statement missing. So if at least one variable is missing on each record then there is no data to model.

A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	V0	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	T1	T2	T3	T4	T5	T6	T7	T8	T9	T10	T11	T12	T13	T14	T15	T16	T17	T18	T19
1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0

A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	V0	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	T1	T2	T3	T4	T5	T6	T7	T8	T9	T10	T11	T12	T13	T14	T15	T16	T17	T18	T19
1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0

A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	V0	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	T1	T2	T3	T4	T5	T6	T7	T8	T9	T10	T11	T12	T13	T14	T15	T16	T17	T18	T19
1	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0		0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0