BookmarkSubscribeRSS Feed
JerryLee
Calcite | Level 5

I am trying to use call NLPNRA to generate a maximum likelihood estimate. IML gives an ERROR message saying I am taking log of zero. But I checked the arguments in all the log commands, none of them is zero. Below is the output of the program and error message. Thanks for help.

11108  %macro infere(rep1,rep2,mcno);

11109

11110  proc datasets library=JM;

11111  delete parajm&rep1;

11112  run;

11113

11114  proc iml;

11115  u=J(1,2,.);

11116  uu=J(1,2,.);

11117  zero=J(1,2,0);

11118  var=I(2);

11119  call randseed(1);

11120  do h1=1 to &mcno;

11121     norma=randnormal(1,zero,var);

11122     do g1=1 to 2;

11123        u[g1]=cdf("NORMAL",norma[g1]);

11124     end;

11125     uu=uu//u;

11126  end;

11127

11128  uu1=uu[2:nrow(uu),];

11129  create unifor from uu1;

11130  append from uu1;

11131  quit;

11132

11133  %do i=&rep1 %to &rep2;

11134

11135  data enable2fact;

11136  set JM.finaldata1;

11137  if rep=&i;

11138  run;

11139

11140  proc sort data=enable2fact;

11141  by ID;

11142  run;

11143

11144  proc means data=enable2fact;

11145  by id;

11146  var tij;

11147  output out=no1;

11148  run;

11149

11150  data no2;

11151  set no1;

11152  rename tij=N;

11153  if _STAT_='N';

11154  keep tij;

11155  run;

11156

11157  proc iml;

11158  *******************************************define log likelihood;

11159  start F_GLOBAL(x);

11160    p=x[1];

11161    lambda1t=x[2];lambda1=x[3];

11162    phi1=x[4];phi2=x[5];phi3=x[6];phi=x[7];

11163    beta=x[,8:11];

11164

11165  use no2;

11166  read all into n;

11167  use enable2fact;

11168  read all var {tij_b} into tijb;

11169  read all var {tij} into tij;

11170  read all var {trt} into trt;

11171  read all var {factpal} into y;

11172  read all var {intercept} into intcep;

11173  read all var {duration_month} into s;

11174  read all var {death} into D;

11175  use unifor;

11176  read all var {col1 col2} into uu;

11177

11178  m=0;lyd=0;

11179

11180  do i=1 to nrow(n);

11181  m=m+n;

11182  si=s;

11183  **********************************log likelihood of mixed model part;

11184  rando=j(n,n,1);

11185  epsi=I(n);

11186  Ri=j(n,n,1);

11187  tib=tijb[m-n+1:m];

11188  ti=tij[m-n+1:m];

11189  *******************************generate Ri;

11190    do j=1 to n;

11191      do k=j to n;

11192        Ri[k,j]=exp(-phi*((abs(tib-tib))**p));

11193        Ri[j,k]=exp(-phi*((abs(tib-tib))**p));

11194      end;

11195    end;

11196

11197  Vi=phi1*rando+phi2*epsi+phi3*Ri;

11198  yi=y[m-n+1:m];

11199  incepti=intcep[m-n+1:m];

11200  trti=trt[m-n+1:m];

11201  one=J(n,1,1);

11202

11203  xib=incepti||trti||tib;

11204  beta1=J(3,1,1);

11205  beta1[1]=beta[1];beta1[2]=trt*beta[2];beta1[3]=beta[3]+trt*beta[4];

11206    if D=1 then do;

11207

11208     ly=-0.5*(log(det(Vi))+(t(yi-xib*beta1))*inv(Vi)*(yi-xib*beta1));

11209     *************************************************************************log likelihood

11209! for observed events;

11210     if trt=1 then ld=log(lambda1t)-lambda1t*si;

11211     if trt=0 then ld=log(lambda1)-lambda1*si;

11212     lyd=lyd+ly+ld;

11213    end;

11214  *************************************************************************log likelihood for

11214! censored events;

11215    if D=0 then do;

11216      ff=0;

11217      do l=1 to nrow(uu);

11218        if trt=0 then surv=(1/lambda1)*log(1/(1-uu[l,1]));

11219        if trt=1 then surv=(1/lambda1t)*log(1/(1-uu[l,2]));

11220        xbeta=(beta[1]+beta[2]*trt)*one+(beta[3]+beta[4]*trt)*one*surv-(beta[3]+beta[4]*t

11220! rt)*ti;

11221        fyd=exp(-0.5*(t(yi-xbeta))*inv(Vi)*(yi-xbeta));

11222        if surv<si then fyd=0;

11223        ff=ff+fyd;

11224      end;

11225

11226      lyd=lyd+log(ff/nrow(uu))-0.5*log(det(Vi));

11227    end;

11228   end;

11229

11230  f=lyd;

11231

11232  return(f);

11233  finish F_GLOBAL;

11234

11235  ******************Making upper and lower bounds for the parameters;

11236  con=J(3,13,.);

11237  boundint=500;

11238  boundbeta=500;

11239  positive=10**(-10);

11240  varianlower=3;

11241  varianupper=500;

11242  boundphi=10;

11243  boundlam=1;

11244  do i=1 to 2;

11245   do j=2 to 11;

11246      if i=1 then do;

11247        if j<=3 then con[i,j]=positive;

11248        if j>=4 & j<=6 then con[i,j]=varianlower;

11249        if j=7 then con[i,j]=positive;

11250        if j=8 then con[i,j]=positive;

11251        if j>=9 then con[i,j]=-boundbeta;

11252      end;

11253      if i=2 then do;

11254        if j<=3 then con[i,j]=boundlam;

11255        if j>=4 & j<=6 then con[i,j]=varianupper;

11256        if j=7 then con[i,j]=boundphi;

11257        if j=8 then con[i,j]=boundint;

11258        if j>=9 then con[i,j]=boundbeta;

11259      end;

11260   end;

11261  end;

11262

11263  con[3,1]=1;con[3,12]=0;con[3,13]=2;

11264

11265  print con;

11266

11267  ********initial value;

11268  x={2 0.04954 0.07649 270.560 118.793 151.389 0.0284 108.702 12.560 1 -0.5};

11269

11270  ********threshold parameters;

11271  opt=J(1,9,.);

11272  opt[1]=1;

11273  opt[2]=5;

11274  tc=J(1,11,.);

11275  ********maximization;

11276  call nlpnra(rc,xres,"F_GLOBAL",x,opt,con,tc);

11277  create solution from xres;

11278  append from xres;

11279  quit;

11280

11281  proc datasets;

11282  append base=JM.parajm&rep1 data=solution;

11283  run;

11284

11285  %end;

11286

11287  %mend;

11288

11289  %infere(rep1=2,rep2=2,mcno=10000);

NOTE: PROCEDURE DATASETS used (Total process time):

      real time           1:50.81

      cpu time            7.31 seconds

NOTE: Writing HTML Body file: sashtml132.htm

                                             Directory

                                  Libref         JM

                                  Engine         V9

                                  Physical Name  D:\lzg\jointmodel

                                  Filename       D:\lzg\jointmodel

                                        Member

              #  Name                   Type     File Size  Last Modified

              1  CICTR                  DATA          5120  15Jul11:10:02:23

              2  CIDIFF                 DATA          5120  15Jul11:10:02:23

              3  CITRT                  DATA          5120  15Jul11:10:02:23

              4  DEATHIMPUT             DATA        263168  04May11:11:19:36

              5  DURATION               DATA          5120  11Mar11:14:29:15

              6  ENABLE050311           DATA        263168  03May11:09:21:53

              7  ENABLE2                DATA        967680  04May11:11:24:16

              8  ENABLE2FACT            DATA        689152  01Mar11:13:53:37

              9  ENABLEIIFORLI12012010  DATA        246784  08Dec10:09:59:54

             10  FINALDATA1             DATA         74752  23Sep11:21:14:31

             11  NO2                    DATA          5120  01Mar11:13:53:37

             12  PARAJM1                DATA          9216  26Sep11:09:29:30

             13  PARAJM2                DATA          9216  26Sep11:10:47:40

             14  PARAJM3                DATA          9216  26Sep11:09:44:16

             15  PARAJMD1               DATA          9216  23Sep11:20:56:28

             16  QALYDIFF1              DATA         16384  14Jul11:19:33:19

             17  QALYDIFF188            DATA         16384  14Jul11:18:45:05

             18  QALYDIFF188N           DATA         16384  09Jul11:10:30:34

             19  QALYDIFF1N             DATA         24576  09Jul11:11:23:09

             20  QALYDIFF251            DATA         24576  05Jul11:23:12:57

             21  QALYDIFF376            DATA         16384  14Jul11:19:07:25

             22  QALYDIFF376N           DATA         16384  09Jul11:10:35:50

             23  QALYDIFF501            DATA         24576  06Jul11:02:31:57

             24  QALYDIFF563            DATA         16384  14Jul11:19:25:50

             25  QALYDIFF563N           DATA         16384  09Jul11:10:58:49

             26  QALYDIFF751            DATA         16384  14Jul11:06:56:01

             27  QALYDIFF751N           DATA         24576  08Jul11:21:12:26

             28  QALYDIFF876            DATA         16384  14Jul11:06:46:01

             29  QALYDIFF876N           DATA         16384  08Jul11:21:50:09

             30  QALYDIFFERENCE         DATA          5120  11Jul11:21:33:27

             31  RESAMP                 DATA     145179648  01Jul11:09:57:14

             32  SAMPLESIZELLAR1        DATA         16384  29Jul11:16:44:20

             33  SAMPLESIZELLCS         DATA         16384  29Jul11:16:42:38

             34  SOLUTION               DATA         17408  24Jun11:12:55:40

             35  UNIFORM                DATA         21504  01Mar11:13:53:37

             36  VISITALL               DATA       4293632  04May11:11:24:16

             37  WITHDRAW               DATA          9216  03May11:09:05:58

NOTE: Deleting JM.PARAJM2 (memtype=DATA).

NOTE: PROCEDURE DATASETS used (Total process time):

      real time           0.06 seconds

      cpu time            0.06 seconds

NOTE: IML Ready

NOTE: Module RANDNORMAL loaded from the storage SASHELP.IMLMLIB.

NOTE: Module ROWVEC loaded from the storage SASHELP.IMLMLIB.

NOTE: Exiting IML.

NOTE: The data set WORK.UNIFOR has 10000 observations and 2 variables.

NOTE: 7746 workspace compresses.

NOTE: PROCEDURE IML used (Total process time):

      real time           0.22 seconds

      cpu time            0.21 seconds

NOTE: There were 467 observations read from the data set JM.FINALDATA1.

NOTE: The data set WORK.ENABLE2FACT has 111 observations and 18 variables.

NOTE: DATA statement used (Total process time):

      real time           0.00 seconds

      cpu time            0.01 seconds

NOTE: There were 111 observations read from the data set WORK.ENABLE2FACT.

NOTE: The data set WORK.ENABLE2FACT has 111 observations and 18 variables.

NOTE: PROCEDURE SORT used (Total process time):

      real time           0.00 seconds

      cpu time            0.00 seconds

NOTE: Writing HTML Body file: sashtml133.htm

NOTE: The above message was for the following BY group:

      id=1

NOTE: There were 111 observations read from the data set WORK.ENABLE2FACT.

NOTE: The data set WORK.NO1 has 100 observations and 5 variables.

NOTE: PROCEDURE MEANS used (Total process time):

      real time           0.07 seconds

      cpu time            0.07 seconds

NOTE: There were 100 observations read from the data set WORK.NO1.

NOTE: The data set WORK.NO2 has 20 observations and 1 variables.

NOTE: DATA statement used (Total process time):

      real time           0.00 seconds

      cpu time            0.00 seconds

NOTE: IML Ready

NOTE: Module F_GLOBAL defined.

NOTE: Writing HTML Body file: sashtml134.htm

ERROR: (execution) Invalid argument to function.

operation : LOG at line 11289 column 1

operands  : _TEM1002

_TEM1002      1 row       1 col     (numeric)

         0

statement : ASSIGN at line 11289 column 1

traceback : module F_GLOBAL at line 11289 column 1

ERROR: Execution error as noted previously. (rc=100)

operation : NLPNRA at line 11289 column 1

operands  : *LIT1116, x, opt, con, tc

*LIT1116      1 row       1 col     (character, size 😎

F_GLOBAL

x      1 row      11 cols    (numeric)

opt      1 row       9 cols    (numeric)

         1         5         .         .         .         .         .         .         .

con      3 rows     13 cols    (numeric)

tc      1 row      11 cols    (numeric)

statement : CALL at line 11289 column 1

ERROR: Matrix xres has not been set to a value.

statement : CREATE at line 11289 column 1

ERROR: No data set is currently open for output.

statement : APPEND at line 11289 column 1

NOTE: Exiting IML.

NOTE: 194882 workspace compresses.

NOTE: The SAS System stopped processing this step because of errors.

NOTE: PROCEDURE IML used (Total process time):

      real time           1:36.72

      cpu time            1:36.40

NOTE: Writing HTML Body file: sashtml135.htm

                                             Directory

           Libref         WORK

           Engine         V9

           Physical Name  C:\Users\Zhigang\AppData\Local\Temp\SAS Temporary Files\_TD2780

           Filename       C:\Users\Zhigang\AppData\Local\Temp\SAS Temporary Files\_TD2780

                                    Member      File

                    #  Name         Type        Size  Last Modified

                    1  AAA          DATA       13312  26Sep11:10:12:46

                    2  ENABLE2FACT  DATA       25600  26Sep11:10:49:31

                    3  FF1          DATA        5120  26Sep11:10:12:46

                    4  FF11         DATA        5120  26Sep11:10:36:27

                    5  FYD          DATA       46080  26Sep11:09:56:52

                    6  FYD11        DATA       46080  26Sep11:09:57:17

                    7  NO1          DATA        9216  26Sep11:10:49:31

                    8  NO2          DATA        5120  26Sep11:10:49:31

                    9  SASMACR      CATALOG     5120  26Sep11:09:15:51

                   10  SOLUTION     DATA        9216  26Sep11:09:44:16

                   11  UNIFOR       DATA      164864  26Sep11:10:49:31

                   12  _FMTDESC     DATA        9216  26Sep11:10:01:38

                   13  _QWFFMT      DATA        9216  26Sep11:10:01:38

NOTE: Appending WORK.SOLUTION to JM.PARAJM2.

NOTE: BASE data set does not exist. DATA file is being copied to BASE file.

NOTE: There were 1 observations read from the data set WORK.SOLUTION.

NOTE: The data set JM.PARAJM2 has 1 observations and 11 variables.

7 REPLIES 7
Rick_SAS
SAS Super FREQ

Because of the macro call, the ERROR message doesn't give any useful information about which LOG call is failing.Two suggestions:

1) For debugging, define %LET i=2; %LET mcno = 10000; and get rid of the other macro statements.

2) You have many LOG statements, but we don't know which temporary argument is going negative.  Go through the code and assign real variables to the arguments:

temp1 = ff/nrow(uu);

temp2 = det(Vi);

etc.

Then run the program. When the program fails you'll get a real line number that points to the error, and you'll get the actual name of a variable that is negative.  For example, if temp1 is negative, you'll get the error message:

ERROR: (execution) Invalid argument to function.

operation : LOG at line XXXX column CCC

operands : TEMP1

TEMP1 1 row 1 col (numeric)

0

Rick

SAS/IML blog: http://blogs.sas.com/content/iml

PS. Unless I'm mistaken, your first call to PROC IML just fills an (&mcno x 2) matrix with uniform random values.  You can just use

uu1=J(&mcno,2,.);

call randgen(uu1, "Uniform");

JerryLee
Calcite | Level 5

Hi Rick,

Thanks a lot for the prompt advice. I tried what you suggested and found that the error message refered to temp1=ff/nrow(uu). To find out which step it is equal to zero, I printed out all the values of temp1. The strange thing is that none of them is zero. And the objective function F_GLOBAL(x) can be calculated the at the initial value. It is equal to -431.3269953. Do you know why IML give this error message? Thanks a lot. Below has the new output of the error message.

12092  proc iml;

NOTE: IML Ready

12093  *******************************************define log likelihood;

12094  start F_GLOBAL(x);

12095    p=x[1];

12096    lambda1t=x[2];

12096!                 lambda1=x[3];

12097    phi1=x[4];

12097!             phi2=x[5];

12097!                       phi3=x[6];

12097!                                 phi=x[7];

12098    beta=x[,8:11];

12099

12100  use no2;

12101  read all into n;

12102  use enable2fact;

12103  read all var {tij_b} into tijb;

12104  read all var {tij} into tij;

12105  read all var {trt} into trt;

12106  read all var {factpal} into y;

12107  read all var {intercept} into intcep;

12108  read all var {duration_month} into s;

12109  read all var {death} into D;

12110  use unifor;

12111  read all var {col1 col2} into uu;

12112

12113  m=0;

12113!     lyd=0;

12114

12115  do i=1 to nrow(n);

12116  m=m+n;

12117  si=s;

12118  **********************************log likelihood of mixed model part;

12119  rando=j(n,n,1);

12120  epsi=I(n);

12121  Ri=j(n,n,1);

12122  tib=tijb[m-n+1:m];

12123  ti=tij[m-n+1:m];

12124  *******************************generate Ri;

12125    do j=1 to n;

12126      do k=j to n;

12127        Ri[k,j]=exp(-phi*((abs(tib-tib))**p));

12128        Ri[j,k]=exp(-phi*((abs(tib-tib))**p));

12129      end;

12130    end;

12131

12132  Vi=phi1*rando+phi2*epsi+phi3*Ri;

12133  yi=y[m-n+1:m];

12134  incepti=intcep[m-n+1:m];

12135  trti=trt[m-n+1:m];

12136  one=J(n,1,1);

12137

12138  xib=incepti||trti||tib;

12139  beta1=J(3,1,1);

12140  beta1[1]=beta[1];

12140!                  beta1[2]=trt*beta[2];

12140!                                          beta1[3]=beta[3]+trt*beta[4];

12141    if D=1 then do;

12142

12143     ly=-0.5*(log(det(Vi))+(t(yi-xib*beta1))*inv(Vi)*(yi-xib*beta1));

12144     *************************************************************************log likelihood

12144! for observed events;

12145     if trt=1 then ld=log(lambda1t)-lambda1t*si;

12146     if trt=0 then ld=log(lambda1)-lambda1*si;

12147     lyd=lyd+ly+ld;

12148    end;

12149  *************************************************************************log likelihood for

12149! censored events;

12150    if D=0 then do;

12151      ff=0;

12152      do l=1 to nrow(uu);

12153        if trt=0 then surv=(1/lambda1)*log(1/(1-uu[l,1]));

12154        if trt=1 then surv=(1/lambda1t)*log(1/(1-uu[l,2]));

12155        xbeta=(beta[1]+beta[2]*trt)*one+(beta[3]+beta[4]*trt)*one*surv-(beta[3]+beta[4]*t

12155! rt)*ti;

12156        fyd=exp(-0.5*(t(yi-xbeta))*inv(Vi)*(yi-xbeta));

12157        if surv<si then fyd=0;

12158        ff=ff+fyd;

12159      end;

12160  temp1=ff/nrow(uu);

12160!                   temp2=det(Vi);

12161      lyd=lyd+log(temp1)-0.5*log(temp2);

12162      print temp1;

12163

12164    end;

12165   end;

12166

12167  f=lyd;

12168

12169  return(f);

12170  finish F_GLOBAL;

NOTE: Module F_GLOBAL defined.

12171

12172  ******************Making upper and lower bounds for the parameters;

12173  con=J(3,13,.);

12174  boundint=500;

12175  boundbeta=500;

12176  positive=10**(-10);

12177  varianlower=3;

12178  varianupper=500;

12179  boundphi=10;

12180  boundlam=1;

12181  do i=1 to 2;

12182   do j=2 to 11;

12183      if i=1 then do;

12184        if j<=3 then con[i,j]=positive;

12185        if j>=4 & j<=6 then con[i,j]=varianlower;

12186        if j=7 then con[i,j]=positive;

12187        if j=8 then con[i,j]=positive;

12188        if j>=9 then con[i,j]=-boundbeta;

12189      end;

12190      if i=2 then do;

12191        if j<=3 then con[i,j]=boundlam;

12192        if j>=4 & j<=6 then con[i,j]=varianupper;

12193        if j=7 then con[i,j]=boundphi;

12194        if j=8 then con[i,j]=boundint;

12195        if j>=9 then con[i,j]=boundbeta;

12196      end;

12197   end;

12198  end;

12199

12200  con[3,1]=1;

12200!            con[3,12]=0;

12200!                        con[3,13]=2;

12201

12202  print con;

NOTE: Writing HTML Body file: sashtml147.htm

12203

12204  ********initial value;

12205  *x={2 0.04954 0.07649 270.560 118.793 151.389 0.0284 108.702 12.560 1 -0.5};

12206  x={2 0.0584913446 0.0810734988 270.5599998 118.79300677 151.38885874 0.0378699563

12206! 108.70197893 12.56070818 0.9993222093 -0.500002042};

12207  xxx=F_GLOBAL(x);

12208  print xxx;

12209  ********threshold parameters;

12210  opt=J(1,9,.);

12211  opt[1]=1;

12212  opt[2]=5;

12213  tc=J(1,11,.);

12214  ********maximization;

12215  call nlpnra(rc,xres,"F_GLOBAL",x,opt,con,tc);

ERROR: (execution) Invalid argument to function.

operation : LOG at line 12161 column 16

operands  : temp1

temp1      1 row       1 col     (numeric)

         0

statement : ASSIGN at line 12161 column 5

traceback : module F_GLOBAL at line 12161 column 5

ERROR: Execution error as noted previously. (rc=100)

operation : NLPNRA at line 12215 column 1

operands  : *LIT1116, x, opt, con, tc

*LIT1116      1 row       1 col     (character, size 😎

F_GLOBAL

x      1 row      11 cols    (numeric)

opt      1 row       9 cols    (numeric)

         1         5         .         .         .         .         .         .         .

con      3 rows     13 cols    (numeric)

tc      1 row      11 cols    (numeric)

statement : CALL at line 12215 column 1

12216  create solution from xres;

ERROR: Matrix xres has not been set to a value.

statement : CREATE at line 12216 column 1

12217  append from xres;

ERROR: No data set is currently open for output.

statement : APPEND at line 12217 column 1

12218  quit;

NOTE: Exiting IML.

NOTE: 190554 workspace compresses.

NOTE: The SAS System stopped processing this step because of errors.

NOTE: PROCEDURE IML used (Total process time):

      real time           1:37.33

      cpu time            1:35.27

Rick_SAS
SAS Super FREQ

I think you have to print temp1 BEFORE you call the LOG function. You are only seeing the result of previous iterations.

I think I might see what is happening. Your loop is essentially as follows:

      ff=0;

      do l=1 to nrow(uu);

        surv= ...;

        if surv<si then fyd=0;

        ff=ff+fyd;

      end;

I think si is large compared to surv so that surv < si is always true.

Of course, the real question is what to do about it, and hopefully you can figure that out.  If the value of Duration_Month (which determines si) is bad, you can fix it. Otherwise, it could be that lambda1t and lambda1 are getting small during the optimization. This would case surv to get large. If this is the problem, you might want to make sure they remain bounded away from zero.

By the way, if you haven't read this paper by Charlie Hallahan on MLE estimation, I recommend it: http://www2.sas.com/proceedings/sugi22/STATS/PAPER293.PDF

JerryLee
Calcite | Level 5

Hi Rick,

Thank you so much! This is really helpful. I got the problem resolved. I was just wondering if IML or "call NLPNRA" also offer the calculation of Fisher information matrix which is a very natural thing associated with maximum likelihood estiamtors (MLEs). Thanks again.

Rick_SAS
SAS Super FREQ

Glad you resolved the problem.

I am not very familiar with the Fisher information matrix, so I'll let someone else offer advice on that.  As far as I know, there are no "built in" routines for computing it, since the NLP functions are designed for general optimization, not for the specific case of MLE.

JerryLee
Calcite | Level 5

Hi Rick,

Do you know if IML has a built in function or command to calculate the second order derivatives of a function? Thanks a lot.

Rick_SAS
SAS Super FREQ

The NLPFDD function uses computes numerical first and second derivatives:

http://blogs.sas.com/content/iml/2011/10/14/hints-for-derivatives/

SAS Innovate 2025: Register Now

Registration is now open for SAS Innovate 2025 , our biggest and most exciting global event of the year! Join us in Orlando, FL, May 6-9.
Sign up by Dec. 31 to get the 2024 rate of just $495.
Register now!

Multiple Linear Regression in SAS

Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.

Find more tutorials on the SAS Users YouTube channel.

From The DO Loop
Want more? Visit our blog for more articles like these.
Discussion stats
  • 7 replies
  • 2634 views
  • 0 likes
  • 2 in conversation