topic Re: DFBETAS in Statistical Procedures

DFBETAS

tanya13 — Mon, 28 May 2018 23:35:11 GMT

HI I AM RUNNING THE FOLLOWING CODE IN ORDER TO COMPUTE THE VALUE OF THE DFBETAS FOR EACH STATE, BUT I AM GETTING MANY WARNINGS AND ERRORS

proc reg data=newsalaries;
model teach_sal = edu_spen geograph area2 area3/ influence vif;
output out=new1 h=hatdiags p=pred r=resid
student=stres cookd= cooksD dffits= dffits ;
title1 'DFBETAS for each state';
symbol1 value=dot pointlabel = (height = 10pt
'#state');
run;
plot stres*pred;
plot hatdiags*pred;
plot cooksD*pred;
run;

THESE ARE THE WARNINGS AND ERRORS I AM GETTING AND I HAVE TRIED REPLACING THE "PRED" WITH "GEOGRAPH" BUT THEN I GET ANOTHER ERROR. I AM USING SAS VERSION 9.4. I WOULD REALLY APPRECIATE SOME HELP WITH THIS. THANK YOU.

551 proc reg data=newsalaries;
552 model teach_sal = edu_spen geograph area2 area3/ influence vif;
ERROR: Variable geograph in list does not match type prescribed for this list.
NOTE: The previous statement has been deleted.
553 output out=new1 h=hatdiags p=pred r=resid
554 student=stres cookd= cooksD dffits= dffits ;
555 title1 'DFBETAS for each state';
556 symbol1 value=dot pointlabel = (height = 10pt
557 '#state');
558 run;

WARNING: No variables specified for an SSCP matrix. Execution terminating.
NOTE: PROCEDURE REG used (Total process time):
real time 0.04 seconds
cpu time 0.03 seconds

NOTE: The data set WORK.NEW1 has 0 observations and 0 variables.
WARNING: Data set WORK.NEW1 was not replaced because new file is incomplete.

559 plot stres*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

560 plot hatdiags*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

561 plot cooksD*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

562 run;

Re: DFBETAS

Reeza — Tue, 29 May 2018 01:03:09 GMT

Fix the errors in the order they appear. When you fix the first ones the latter ones will likely go away too.

So your first error:

ERROR: Variable geograph in list does not match type prescribed for this list.

So GeoGraph is likely a character variable when it should be numeric to be included in a regression model for obvious reasons.

Convert the variable to a numeric variable and it will work.

Your PLOT statements later on don't make sense though, they aren't in a PROC. Wherever you're referencing is also very outdated, PLOT statements within PROC REG have been disabled for a while and PROC GPLOT is outdated as well. If you want to create a graph try SGPLOT procedures instead. Make sure to reference the correct documentation for your version of SAS/STAT which you can check using:

proc product_status;run;

The latest version is SAS/STAT 14.3

@tanya13 wrote:

HI I AM RUNNING THE FOLLOWING CODE IN ORDER TO COMPUTE THE VALUE OF THE DFBETAS FOR EACH STATE, BUT I AM GETTING MANY WARNINGS AND ERRORS

proc reg data=newsalaries;
model teach_sal = edu_spen geograph area2 area3/ influence vif;
output out=new1 h=hatdiags p=pred r=resid
student=stres cookd= cooksD dffits= dffits ;
title1 'DFBETAS for each state';
symbol1 value=dot pointlabel = (height = 10pt
'#state');
run;
plot stres*pred;
plot hatdiags*pred;
plot cooksD*pred;
run;

THESE ARE THE WARNINGS AND ERRORS I AM GETTING AND I HAVE TRIED REPLACING THE "PRED" WITH "GEOGRAPH" BUT THEN I GET ANOTHER ERROR. I AM USING SAS VERSION 9.4. I WOULD REALLY APPRECIATE SOME HELP WITH THIS. THANK YOU.

551 proc reg data=newsalaries;
552 model teach_sal = edu_spen geograph area2 area3/ influence vif;
ERROR: Variable geograph in list does not match type prescribed for this list.
NOTE: The previous statement has been deleted.
553 output out=new1 h=hatdiags p=pred r=resid
554 student=stres cookd= cooksD dffits= dffits ;
555 title1 'DFBETAS for each state';
556 symbol1 value=dot pointlabel = (height = 10pt
557 '#state');
558 run;

WARNING: No variables specified for an SSCP matrix. Execution terminating.
NOTE: PROCEDURE REG used (Total process time):
real time 0.04 seconds
cpu time 0.03 seconds

NOTE: The data set WORK.NEW1 has 0 observations and 0 variables.
WARNING: Data set WORK.NEW1 was not replaced because new file is incomplete.

559 plot stres*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

560 plot hatdiags*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

561 plot cooksD*pred;
----
180

ERROR 180-322: Statement is not valid or it is used out of proper order.

562 run;

Re: DFBETAS

PaigeMiller — Tue, 29 May 2018 01:22:58 GMT

Converting GEOGRAPH to numeric may not make sense, if the variable is not something that can be converted to numbers (like if it is a US state, then numbers don't make sense). In that case PROC GLM could be used to handle the model fit but it won't compute DFBETAS. Yes, I know, that's not fair.

You could create dummy variables and use PROC REG and computer DFBETAS.