Help using Base SAS procedures

Standardized 95%CI for Standardized estimates

Accepted Solution Solved
Reply
Contributor
Posts: 37
Accepted Solution

Standardized 95%CI for Standardized estimates

Dear all, thank you very much for all your support.

 

I have a new problem.

 

Now I am doing some linear regression, I am able to get standardized estimates but not standardized 95% CI intervals.  I am getting 95% CI but not standardized. Please help.

 

Here are my codes 

 proc reg data=ALICEPAPER1MAY2018;
  model log_WHtR_2005=PPPL1Satfat PPPL2N3VLCAA PPPL3n62dbond PPPL4DGLAadreDPA PPPL5MUFA PPPL6ALA/ ss1 ss2 stb clb covb corrb ;
   run;

here is the output 

SAS Output

Parameter EstimatesVariable Label DF ParameterEstimate StandardError t Value Pr > |t| Type I SS Type II SS StandardizedEstimate 95% Confidence LimitsIntercept1PPPL1Satfat1PPPL2N3VLCAA1PPPL3n62dbond1PPPL4DGLAadreDPA1PPPL5MUFA1PPPL6ALA1
Intercept-0.704430.00577-122.10<.0001350.03800350.299480-0.71576-0.69310
 0.051310.005798.86<.00011.852431.845200.306580.039940.06268
 0.035840.005766.22<.00010.910550.909780.215270.024530.04714
 -0.006140.00577-1.060.28740.026480.02663-0.03683-0.017460.00518
 0.011530.005762.000.04560.094090.094200.069270.000224080.02285
 -0.015110.00576-2.620.00890.161720.16149-0.09069-0.02643-0.00379
 -0.015190.00578-2.630.00880.162410.16241-0.09096-0.02653-0.00385

 

Please help, my supervisor insist we have to generate that, I have tried a few codes, that I have found online, without success, some of them are 

  proc standard data=ALICEPAPER1MAY2018 mean=0 std=1 out=AlICEMAY_2018;
  proc reg; 
model log_WHtR_2005=PPPL1Satfat PPPL2N3VLCAA PPPL3n62dbond PPPL4DGLAadreDPA PPPL5MUFA PPPL6ALA /stb;
   run;

I am only getting standardized estimates with this code. Please help me get standardized 95%CI for standardized estimates.

Thank you. I am looking forward to hearing from you.

 

very kind regards

Achieng


Accepted Solutions
Solution
‎05-18-2018 05:55 AM
SAS Super FREQ
Posts: 4,239

Re: Standardized 95%CI for Standardized estimates

You can use PROC STDIZE to physically standardize the variables and then request CLB for the same model on the standardized variables:

 

/* original parameter estimates, including STB */
proc reg data=sashelp.class plots=none;
  model height=weight age/ stb clb;
  ods select ParameterEstimates;
quit;

/* standardize the numerical data */
proc stdize data=sashelp.class out=stdClass;
run;

/* regression on the standardized data gives same parameter 
   estimates as the STB option on the original data. But now
   the CLB option gives the CIs for the standardized coefficients */
proc reg data=stdClass plots=none;
  model height=weight age/ clb;
  ods select ParameterEstimates;
quit;

View solution in original post


All Replies
Super User
Posts: 23,683

Re: Standardized 95%CI for Standardized estimates

[ Edited ]

Once the variables are standardized, the 95% CI is on the standardized data. 

The 95% CI on the estimates is from the PROC REG statement using the OUTEST table.

 

 proc reg data=ALICEPAPER1MAY2018 OUTEST = WANT;

Then check the WANT data set for the estimates.

 


@Achieng wrote:

Dear all, thank you very much for all your support.

 

I have a new problem.

 

Now I am doing some linear regression, I am able to get standardized estimates but not standardized 95% CI intervals.  I am getting 95% CI but not standardized. Please help.

 

Here are my codes 

 proc reg data=ALICEPAPER1MAY2018;
  model log_WHtR_2005=PPPL1Satfat PPPL2N3VLCAA PPPL3n62dbond PPPL4DGLAadreDPA PPPL5MUFA PPPL6ALA/ ss1 ss2 stb clb covb corrb ;
   run;

here is the output 

SAS Output

Parameter EstimatesVariable Label DF ParameterEstimate StandardError t Value Pr > |t| Type I SS Type II SS StandardizedEstimate 95% Confidence LimitsIntercept1PPPL1Satfat1PPPL2N3VLCAA1PPPL3n62dbond1PPPL4DGLAadreDPA1PPPL5MUFA1PPPL6ALA1
Intercept -0.70443 0.00577 -122.10 <.0001 350.03800 350.29948 0 -0.71576 -0.69310
  0.05131 0.00579 8.86 <.0001 1.85243 1.84520 0.30658 0.03994 0.06268
  0.03584 0.00576 6.22 <.0001 0.91055 0.90978 0.21527 0.02453 0.04714
  -0.00614 0.00577 -1.06 0.2874 0.02648 0.02663 -0.03683 -0.01746 0.00518
  0.01153 0.00576 2.00 0.0456 0.09409 0.09420 0.06927 0.00022408 0.02285
  -0.01511 0.00576 -2.62 0.0089 0.16172 0.16149 -0.09069 -0.02643 -0.00379
  -0.01519 0.00578 -2.63 0.0088 0.16241 0.16241 -0.09096 -0.02653 -0.00385

 

Please help, my supervisor insist we have to generate that, I have tried a few codes, that I have found online, without success, some of them are 

  proc standard data=ALICEPAPER1MAY2018 mean=0 std=1 out=AlICEMAY_2018;
  proc reg; 
model log_WHtR_2005=PPPL1Satfat PPPL2N3VLCAA PPPL3n62dbond PPPL4DGLAadreDPA PPPL5MUFA PPPL6ALA /stb;
   run;

I am only getting standardized estimates with this code. Please help me get standardized 95%CI for standardized estimates.

Thank you. I am looking forward to hearing from you.

 

very kind regards

Achieng


 

Solution
‎05-18-2018 05:55 AM
SAS Super FREQ
Posts: 4,239

Re: Standardized 95%CI for Standardized estimates

You can use PROC STDIZE to physically standardize the variables and then request CLB for the same model on the standardized variables:

 

/* original parameter estimates, including STB */
proc reg data=sashelp.class plots=none;
  model height=weight age/ stb clb;
  ods select ParameterEstimates;
quit;

/* standardize the numerical data */
proc stdize data=sashelp.class out=stdClass;
run;

/* regression on the standardized data gives same parameter 
   estimates as the STB option on the original data. But now
   the CLB option gives the CIs for the standardized coefficients */
proc reg data=stdClass plots=none;
  model height=weight age/ clb;
  ods select ParameterEstimates;
quit;
Contributor
Posts: 37

Re: Standardized 95%CI for Standardized estimates

Dear Rick. Thank you.
Let me try this and will get to you As soon as possible.

Regards

Contributor
Posts: 37

Re: Standardized 95%CI for Standardized estimates

Thank you very much.

It worked.

 

 

I love the SAS community.

 

YeahSmiley Wink

 

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 176 views
  • 1 like
  • 3 in conversation