From 129536af79f75d81135de434785196cf82728d66 Mon Sep 17 00:00:00 2001 From: "N. Brouard" Date: Tue, 5 Apr 2022 21:03:51 +0000 Subject: [PATCH] Summary: Fixed quantitative covariates Fixed covariates (dummy or quantitative) with missing values have never been allowed but are ERRORS and program quits. Standard deviations of fixed covariates were wrongly computed. Mean and standard deviations of time varying covariates are still not computed. --- src/imach.c | 52 ++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 40 insertions(+), 12 deletions(-) diff --git a/src/imach.c b/src/imach.c index 7d410a2..549b0a7 100644 --- a/src/imach.c +++ b/src/imach.c @@ -1,6 +1,12 @@ /* $Id$ $State$ $Log$ + Revision 1.310 2022/03/17 08:45:53 brouard + Summary: 99r25 + + Improving detection of errors: result lines should be compatible with + the model. + Revision 1.309 2021/05/20 12:39:14 brouard Summary: Version 0.99r24 @@ -3841,7 +3847,7 @@ double funcone( double *x) /* Fixed */ /* for (k=1; k<=cptcovn;k++) cov[2+nagesqr+k]=covar[Tvar[k]][i]; */ /* for (k=1; k<=ncoveff;k++){ /\* Simple and product fixed Dummy covariates without age* products *\/ */ - for (k=1; k<=ncovf;k++){ /* Simple and product fixed covariates without age* products */ + for (k=1; k<=ncovf;k++){ /* Simple and product fixed covariates without age* products *//* Missing values are set to -1 but should be dropped */ cov[ioffset+TvarFind[k]]=covar[Tvar[TvarFind[k]]][i];/* V5+V4+V3+V4*V3+V5*age+V2+V1*V2+V1*age+V1, only V1 is fixed (k=6)*/ /* cov[ioffset+TvarFind[1]]=covar[Tvar[TvarFind[1]]][i]; */ /* cov[2+6]=covar[Tvar[6]][i]; */ @@ -4696,10 +4702,13 @@ Title=%s
Datafile=%s Firstpass=%d Lastpass=%d Stepm=%d Weight=%d Model=1+age if(s[m][iind]==-1) printf(" num=%ld m=%d, iind=%d s1=%d s2=%d agev at m=%d agebegin=%.2f ageend=%.2f, agemed=%d\n", num[iind], m, iind,s[m][iind],s[m+1][iind], (int)agev[m][iind],agebegin, ageend, (int)((agebegin+ageend)/2.)); freq[s[m][iind]][s[m+1][iind]][(int)agev[m][iind]] += weight[iind]; /* At age of beginning of transition, where status is known */ - for (z1=1; z1<= nqfveff; z1++) { /* Quantitative variables, calculating mean */ - idq[z1]=idq[z1]+weight[iind]; - meanq[z1]+=covar[ncovcol+z1][iind]*weight[iind]; /* Computes mean of quantitative with selected filter */ - stdq[z1]+=covar[ncovcol+z1][iind]*covar[ncovcol+z1][iind]*weight[iind]*weight[iind]; /* *weight[iind];*/ /* Computes mean of quantitative with selected filter */ + for (z1=1; z1<= nqfveff; z1++) { /* Quantitative variables, calculating mean on known values only */ + if(!isnan(covar[ncovcol+z1][iind])){ + idq[z1]=idq[z1]+weight[iind]; + meanq[z1]+=covar[ncovcol+z1][iind]*weight[iind]; /* Computes mean of quantitative with selected filter */ + /* stdq[z1]+=covar[ncovcol+z1][iind]*covar[ncovcol+z1][iind]*weight[iind]*weight[iind]; *//*error*/ + stdq[z1]+=covar[ncovcol+z1][iind]*covar[ncovcol+z1][iind]*weight[iind]; /* *weight[iind];*/ /* Computes mean of quantitative with selected filter */ + } } /* if((int)agev[m][iind] == 55) */ /* printf("j=%d, j1=%d Age %d, iind=%d, num=%09ld m=%d\n",j,j1,(int)agev[m][iind],iind, num[iind],m); */ @@ -4762,16 +4771,19 @@ Title=%s
Datafile=%s Firstpass=%d Lastpass=%d Stepm=%d Weight=%d Model=1+age Printing means of quantitative variables if any */ for (z1=1; z1<= nqfveff; z1++) { - fprintf(ficlog,"Mean of fixed quantitative variable V%d on %.0f individuals sum=%f", ncovcol+z1, idq[z1], meanq[z1]); - fprintf(ficlog,", mean=%.3g\n",meanq[z1]/idq[z1]); + fprintf(ficlog,"Mean of fixed quantitative variable V%d on %.3g (weighted) individuals sum=%f", ncovcol+z1, idq[z1], meanq[z1]); + fprintf(ficlog,", mean=%.3g",meanq[z1]/idq[z1]," stdeviation=%.3g\n",stdq[z1]); if(weightopt==1){ printf(" Weighted mean and standard deviation of"); fprintf(ficlog," Weighted mean and standard deviation of"); fprintf(ficresphtmfr," Weighted mean and standard deviation of"); } - printf(" fixed quantitative variable V%d on %.0f representatives of the population : %6.3g (%6.3g)\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt((stdq[z1]-meanq[z1]*meanq[z1]/idq[z1])/idq[z1])); - fprintf(ficlog," fixed quantitative variable V%d on %.0f representatives of the population : %6.3g (%6.3g)\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt((stdq[z1]-meanq[z1]*meanq[z1]/idq[z1])/idq[z1])); - fprintf(ficresphtmfr," fixed quantitative variable V%d on %.0f representatives of the population : %6.3g (%6.3g)

\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt((stdq[z1]-meanq[z1]*meanq[z1]/idq[z1])/idq[z1])); + /* mu = \frac{w x}{\sum w} + var = \frac{\sum w (x-mu)^2}{\sum w} = \frac{w x^2}{\sum w} - mu^2 + */ + printf(" fixed quantitative variable V%d on %.3g (weighted) representatives of the population : %8.5g (%8.5g)\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt(stdq[z1]/idq[z1]-meanq[z1]*meanq[z1]/idq[z1]/idq[z1])); + fprintf(ficlog," fixed quantitative variable V%d on %.3g (weighted) representatives of the population : %8.5g (%8.5g)\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt(stdq[z1]/idq[z1]-meanq[z1]*meanq[z1]/idq[z1]/idq[z1])); + fprintf(ficresphtmfr," fixed quantitative variable V%d on %.3g (weighted) representatives of the population : %8.5g (%8.5g)

\n", ncovcol+z1, idq[z1],meanq[z1]/idq[z1], sqrt(stdq[z1]/idq[z1]-meanq[z1]*meanq[z1]/idq[z1]/idq[z1])); } /* for (z1=1; z1<= nqtveff; z1++) { */ /* for(m=1;m<=lastpass;m++){ */ @@ -5488,6 +5500,8 @@ void concatwav(int wav[], int **dh, int **bh, int **mw, int **s, double *agedc if(Dummy[k]==0 && Typevar[k] !=1){ /* Dummy covariate and not age product */ switch(Fixed[k]) { case 0: /* Testing on fixed dummy covariate, simple or product of fixed */ + modmaxcovj=0; + modmincovj=0; for (i=1; i<=imx; i++) { /* Loop on individuals: reads the data file to get the maximum value of the modality of this covariate Vj*/ ij=(int)(covar[Tvar[k]][i]); /* ij=0 or 1 or -1. Value of the covariate Tvar[j] for individual i @@ -5501,8 +5515,10 @@ void concatwav(int wav[], int **dh, int **bh, int **mw, int **s, double *agedc else if (ij < modmincovj) modmincovj=ij; if (ij <0 || ij >1 ){ - printf("Information, IMaCh doesn't treat covariate with missing values (-1), individual %d will be skipped.\n",i); - fprintf(ficlog,"Information, currently IMaCh doesn't treat covariate with missing values (-1), individual %d will be skipped.\n",i); + printf("ERROR, IMaCh doesn't treat covariate with missing values V%d=-1, individual %d will be skipped.\n",Tvar[k],i); + fprintf(ficlog,"ERROR, currently IMaCh doesn't treat covariate with missing values V%d=-1, individual %d will be skipped.\n",Tvar[k],i); + fflush(ficlog); + exit(1); } if ((ij < -1) || (ij > NCOVMAX)){ printf( "Error: minimal is less than -1 or maximal is bigger than %d. Exiting. \n", NCOVMAX ); @@ -5577,6 +5593,16 @@ void concatwav(int wav[], int **dh, int **bh, int **mw, int **s, double *agedc break; } /* end switch */ } /* end dummy test */ + if(Dummy[k]==1 && Typevar[k] !=1){ /* Dummy covariate and not age product */ + for (i=1; i<=imx; i++) { /* Loop on individuals: reads the data file to get the maximum value of the modality of this covariate Vj*/ + if(isnan(covar[Tvar[k]][i])){ + printf("ERROR, IMaCh doesn't treat fixed quantitative covariate with missing values V%d=., individual %d will be skipped.\n",Tvar[k],i); + fprintf(ficlog,"ERROR, currently IMaCh doesn't treat covariate with missing values V%d=., individual %d will be skipped.\n",Tvar[k],i); + fflush(ficlog); + exit(1); + } + } + } } /* end of loop on model-covariate k. nbcode[Tvark][1]=-1, nbcode[Tvark][1]=0 and nbcode[Tvark][2]=1 sets the value of covariate k*/ for (k=-1; k< maxncov; k++) Ndum[k]=0; @@ -9501,6 +9527,8 @@ int readdata(char datafile[], int firstobs, int lastobs, int *imax) cutv(stra, strb, line, ' '); if(strb[0]=='.') { /* Missing value */ lval=-1; + coqvar[iv][i]=NAN; + covar[ncovcol+iv][i]=NAN; /* including qvar in standard covar for performance reasons */ }else{ errno=0; /* what_kind_of_number(strb); */ -- 2.43.0