General SAS Programming

[PeteCthulu]

I'm using SAS 9.4.

I have data for about 50 participants, who each read a series of sentences that have the same structure. The reading time (in milliseconds) for each word in each sentence was recorded. I want to get a regression line for how the word length (in characters) relates to reading time for each participant, and I want to then convert the reading time measurement into the corresponding residual from these regression equations for each data point. I've tried to do this with some stupid things, like below. I don't quite get it.  I give an example of how the data look for these variables further below. Please help!

PROC GLM;

model Readingtime= Wordlength;

Do by Participant#;

Participant#	Readingtime	Wordlength	Residual
300	1042	6	?
300	742	4	?
300	615	4	?
300	668	5	?
300	562	3	?
301	677	6	?
301	694	4	?
301	730	4	?
301	512	5	?
301	476	3	?
302	1001	6	?
302	890	4	?
302	640	4	?
302	708	5	?
302	544	3	?

[Reeza · Accepted solution]

proc reg data=sashelp.class;
by sex;
model weight = height age;
output out=want pred = p residual=r;
quit;

proc print data=want;run;

[Reeza]

proc reg data=sashelp.class;
model weight = height age;
output out=want pred = p residual=r;
quit;

proc print data=want;run;

You need to create an OUTPUT data set, see the code above for a fully worked example.

In general, you should always have a DATA statement so that you know which data set your procedure is using. 

[PeteCthulu]

I don't understand how this would get the Yhat for the regression equation for each participant and then get the residual for each data point....

[PaigeMiller]

You need to add

by participant;

into the PROC REG commands.

[Reeza · Accepted solution]

proc reg data=sashelp.class;
by sex;
model weight = height age;
output out=want pred = p residual=r;
quit;

proc print data=want;run;

[PGStats]

A power model would fit pretty well I bet

data temp;
set have;
lReadingTime = log(readingtime);
lWordLength = log(WordLength);
run;

proc mixed data=temp;
class participant;
model lreadingtime = lwordlength / s outpred=outpred residual;
random participant / solution;
ods output SolutionR=SolutionR;
run;

make sure you include all the same words for every reader.

[PeteCthulu]

Thanks! I'll try this too!