two minor bugs that were causing occasionally crazy anderson-darling results

src/BayesWavePost.c

@@ -3361,8 +3361,8 @@ void anderson_darling_test(char *outdir, int ifo, double fmin, double Tobs, char
     mu_re += re;
     mu_im += im;
   }
-  mu_re/=(double)(N/2);
-  mu_im/=(double)(N/2);
+  mu_re/=(double)(Nsamp/2);
+  mu_im/=(double)(Nsamp/2);
   
   //get covariances
   for(n=0; n<Nsamp/2; n++)
@@ -3482,6 +3482,7 @@ void anderson_darling_test(char *outdir, int ifo, double fmin, double Tobs, char
   // p-value of A^2
   double p = 1.-AD(Nsamp,A);
 
+  free(samples);
   sprintf(filename,"%s/%s_anderson_darling_%s.dat",outdir,model,ifos[ifo]);
   FILE *afile = fopen(filename,"w");
   fprintf(afile,"%lg %lg\n",A, p);