topic Parameter estimates by group - can I output them all in one table? in Statistical Procedures

Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 19:33:36 GMT

Hello everyone,

I'm running a multiple regression by group; basically, I have a long list of stock trade data for various companies organized by ticker symbol and a regression is being run by symbol. The result is a set of parameter estimates for each company. When it's time to copy these parameter estimates into excel, it takes me ages because each set of parameter estimates is in a table on its own and i have to manually copy + paste the contents one by one. Is there anyway to have all parameter estimates appear in one table by ticker symbol? Something like this:

Ticker Symbol	(Parameter estimate A)	(Parameter estimate B)
ABC	0.002	12
BBB	0.013	4
D	0.004	14
XYZ	0.079	9

Since a lot of the parameter estimates have several zeros after the decimal (for example: 0.0000002331), I need to make sure the table provides enough decimal places for the estimates (in other words, I don't want it to be reported as "0.00").

Any help is greatly appreciated. As a SAS beginner, I must say I have learnt a great deal from these boards. Thanks in advance.

Re: Parameter estimates by group - can I output them all in one table?

art297 — Wed, 20 Feb 2013 19:48:24 GMT

Would you get what you need by using the outtest option (see: SAS/STAT(R) 9.2 User's Guide, Second Edition )? Otherwise, you should be able to get them from the ODS tables.

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 21:09:04 GMT

Thanks Arthur. I was able to get the table I needed; however, the table still rounds values. For example 2.767E-08 is coming up as 0.000000028. I don't know how to keep them in their original scientific format.

Re: Parameter estimates by group - can I output them all in one table?

Reeza — Wed, 20 Feb 2013 21:26:41 GMT

The format is probably just rounded, try applying a different format to the variables of interest.

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 21:49:32 GMT

In the original output, where the parameter estimates are reported by symbol in individual tables, there is no rounding. When I create that new table they get rounded. I wonder if there is a way I can specify no rounding in the outest function.

Re: Parameter estimates by group - can I output them all in one table?

Reeza — Wed, 20 Feb 2013 21:53:53 GMT

I don't think you're talking about 'rounding', I think you're talking about the format applied to the data.

I don't know of a way to change the format in the output step. I'd change it in a second step or where I was going to use or manipulate the format.

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 21:56:06 GMT

I think it is being rounded because when I run the proc reg step without the outest option the parameter estimate for the first symbol comes up as 2.767E-08 in the output. When I throw in the outest, the new table that contains all the symbols and their parameters is showing 0.000000028 for that same parameter estimate.

Re: Parameter estimates by group - can I output them all in one table?

Reeza — Wed, 20 Feb 2013 22:01:04 GMT

What happens when you do the following. Replace the have with the dataset you created from outest and the parameters with the names of your variables that you think are 'rounded'.

data want;

set have;

format parameters e10.;

run;

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 22:30:33 GMT

Hmm, it didn't change anything.

This is the original table i get after running proc reg:

Parameter Estimates AMD
Variable	DF	Parameter Estimate	Standard Error	t Value	Pr > \|t\|
signchange	1	0.01383	0.00003093	447.04	<.0001
LAGQV	1	1.40022E-07	3.65E-09	38.39	<.0001
sign	1	0.00168	0.00003847	43.74	<.0001
QV	1	2.767E-08	5.03E-09	5.5	<.0001

This is the first row of the table i get from the OUTEST (ive underlined the parameter estimates that correspond to the table above):

Obs	symbol	_MODEL_	_TYPE_	_DEPVAR_	_RMSE_	signchange	LAGQV	sign	QV	Pchange	parameters
1	AMD	MODEL1	PARMS	Pchange	0.02909	0.01383	1.4E-07	0.00168	2.8E-08	-1	.

As you can see, it's still rounding. Did I apply your code properly? The codes that I used are:

PROC REG data=GH2 outest=est;
by symbol;
model Pchange = signchange LAGQV sign QV / noint;

data FINAL;
set est;
format parameters e10.;
run;

proc print data = FINAL;
run;
quit;

Re: Parameter estimates by group - can I output them all in one table?

Reeza — Wed, 20 Feb 2013 22:35:00 GMT

No you didn't. See the part about changing the parameters line.

data FINAL;

set est;

format LAGQV QV e10.;

run;

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 22:35:33 GMT

Woops I left out the parameters that time, but I just ran it again with the parameters and it does brings up the exact same rounded values, but instead of bringing them in number format it brings them up (still rounded) in scientific format.

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 22:37:30 GMT

With the parameters corrected... still rounded but in scientific format.

Obs	symbol	_MODEL_	_TYPE_	_DEPVAR_	_RMSE_	signchange	LAGQV	sign	QV	Pchange
1	AMD	MODEL1	PARMS	Pchange	0.02909	1.38E-02	1.40E-07	1.68E-03	2.77E-08	-1

Re: Parameter estimates by group - can I output them all in one table?

Reeza — Wed, 20 Feb 2013 22:40:29 GMT

2.77 isn't rounded, change the format if you want to see more decimal places.

ie e12. or something.

Re: Parameter estimates by group - can I output them all in one table?

RazzleBayker — Wed, 20 Feb 2013 22:43:38 GMT

Amazing. Sorry for not realizing that :smileyblush:.

Thanks so much your help Reeza and Arthur, I really appreciate it!

-Razzle