1 / 13

Lab 3 Type I, II Error, Sample Size, and Power Henian Chen, M.D., Ph.D.

Lab 3 Type I, II Error, Sample Size, and Power Henian Chen, M.D., Ph.D. Sample 1 n1=10. POPULATION N (5,1). Sample 2 n2=10. Type I Error Believing that some zero population effect is nonzero on the basis of a test of statistical significance. SAS Program for Type 1 Error

shadow
Télécharger la présentation

Lab 3 Type I, II Error, Sample Size, and Power Henian Chen, M.D., Ph.D.

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Lab 3 Type I, II Error, Sample Size, and Power Henian Chen, M.D., Ph.D. Applied Epidemiologic Analysis - P8400 Fall 2002

  2. Sample 1 n1=10 POPULATION N (5,1) Sample 2 n2=10 Type I Error Believing that some zero population effect is nonzero on the basis of a test of statistical significance Applied Epidemiologic Analysis - P8400 Fall 2002

  3. SAS Program for Type 1 Error data type1error; do j=1to2; /* create two samples */ do i=1to10; /* the sample size is 10 */ x=rannor(0)+5; /* random sampling from a normal distribution population, N(5,1)*/ output; end; end; procttest; /* perform independent t-test */ class j; var x; run; Applied Epidemiologic Analysis - P8400 Fall 2002

  4. DATA: type1error Obs j i x 1 1 1 5.34655 2 1 2 4.38660 3 1 3 3.09380 4 1 4 5.24584 5 1 5 5.37742 6 1 6 3.25170 7 1 7 3.57868 8 1 8 4.54776 9 1 9 4.97216 10 1 10 5.03652 11 2 1 6.49084 12 2 2 3.94007 13 2 3 3.84986 14 2 4 4.56861 15 2 5 3.91149 16 2 6 5.25983 17 2 7 5.12068 18 2 8 6.70596 19 2 9 5.00503 20 2 10 5.30035 Applied Epidemiologic Analysis - P8400 Fall 2002

  5. Output 1 of T-Test Statistics Variable j N Mean Std Dev Std Err x 1 10 4.4837 0.8788 0.2779 x 2 10 5.0153 1.0067 0.3183 x Diff (1-2) -0.532 0.9449 0.4226 T-Tests Variable Method Variances DF t Value Pr > |t| x Pooled Equal 18 -1.26 0.2245 x Satterthwaite Unequal 17.7 -1.26 0.2248 Equality of Variances Variable Method Num DF Den DF F Value Pr > F x Folded F 9 9 1.31 0.6923 Applied Epidemiologic Analysis - P8400 Fall 2002

  6. Output 2 of T-Test Statistics Variable j N Mean Std Dev Std Err x 1 10 5.4088 0.7262 0.2296 x 2 10 4.3767 1.0543 0.3334 x Diff (1-2) -0.532 0.9052 0.4048 T-Tests Variable Method Variances DF t Value Pr > |t| x Pooled Equal 18 2.55 0.0201 x Satterthwaite Unequal 16 2.55 0.0214 Equality of Variances Variable Method Num DF Den DF F Value Pr > F x Folded F 9 9 2.11 0.2819 Applied Epidemiologic Analysis - P8400 Fall 2002

  7. POPULATION 1 N (6,1) Sample 1 n1=10 POPULATION 2 N (8,1) Sample 2 n2=10 Type II Error Believing that some nonzero population effect is zero on the basis of a test of statistical significance Applied Epidemiologic Analysis - P8400 Fall 2002

  8. SAS Program for Type 2 Error data type2error; do i=1to10; /* the sample size is 10 */ j=1; /* for sample 1 */ x=rannor(0)+6; /* random sampling from a normal distribution population, N (6,1)*/ output; end; do i=1to10; /* the sample size is 10 */ j=2; /* for sample 2 */ x=rannor(0)+8; /* random sampling from a normal distribution population, N (8,1)*/ output; end; procttest; /* perform independent t-test */ class j; var x; run; Applied Epidemiologic Analysis - P8400 Fall 2002

  9. Output 1 of T-Test Statistics Variable j N Mean Std Dev Std Err x 1 10 6.4906 0.8872 0.2806 x 2 10 7.7403 0.9913 0.3135 x Diff (1-2) -0.532 0.9407 0.4207 T-Tests Variable Method Variances DF t Value Pr > |t| x Pooled Equal 18 -2.97 0.0082 x Satterthwaite Unequal 17.8 -2.97 0.0083 Equality of Variances Variable Method Num DF Den DF F Value Pr > F x Folded F 9 9 1.25 0.7465 Applied Epidemiologic Analysis - P8400 Fall 2002

  10. Output 2 of T-Test Statistics Variable j N Mean Std Dev Std Err x 1 10 6.5582 1.3303 0.4207 x 2 10 7.3892 1.0052 0.3179 x Diff (1-2) -0.8310 1.1790 0.5273 T-Tests Variable Method Variances DF t Value Pr > |t| x Pooled Equal 18 -1.58 0.1325 x Satterthwaite Unequal 16.8 -1.58 0.1337 Equality of Variances Variable Method Num DF Den DF F Value Pr > F x Folded F 9 9 1.75 0.4165 Applied Epidemiologic Analysis - P8400 Fall 2002

  11. Power The probability of rejecting a null hypothesis that is false to a specified degree for a given sample size, Type I error, and effect size Applied Epidemiologic Analysis - P8400 Fall 2002

  12. Increase Alpha Given Sample size Given Effect size Increase Power Increase Sample Size Given Alpha Given Effect size Increase Power Increase Effect Size Given Sample size Given Alpha Increase Power Sample size Effect size Power Alpha Applied Epidemiologic Analysis - P8400 Fall 2002

  13. SAS Program for Power of Case-Control Study data power; c=3.89; /* case:control ratio=200/778 = 3.89 */ n=200; /* number of cases=200 */ do p0=0.05to0.50by0.05; /* exposure prevalence */ do or=1.5to3.0by0.5; /* odds ratios from 1.5 to 3.0 */ q0=1-p0; p1=(p0*or)/(1+p0*(or-1)); q1=1-p1; pbar=(p1+c*p0)/(1+c); qbar=1-pbar; zbeta=(sqrt(n*(p1-p0)**2)-1.96*sqrt((1+1/c)*pbar*qbar)) / sqrt((p1*q1)+p0*q0/c); /* zbeta is the standardized deviate*/ power=probnorm(zbeta); /* PROBNORM returns the probability from the standard normal distribution */ output; end; end; run; procprintdata=power; var p0 or zbeta power; run; Applied Epidemiologic Analysis - P8400 Fall 2002

More Related