G Vishwanath

2012-Jan-03 04:25 UTC

### [R] nls and rbinom function: step factor 0.000488281 reduced below 'minFactor' of 0.000976562

I am trying to learn nls using a simple simulation. I assumed that the binomial prob varies linearly as 0.2 + 0.3*x in x {0,1}, and the objective is to recover the known parameters a=0.2, b=0.3 ..data frame d has 1000 rows... d$x<-runif(0,1) d$y<-rbinom(1000,1,0.2+0.3*d$x) table(d$y,cut(d$x,breaks=5)); (-0.000585,0.199] (0.199,0.399] (0.399,0.599] (0.599,0.799] (0.799,0.999] 0 154 149 130 122 114 1 34 48 71 76 102 z <- nls(y ~ rbinom(1000,1,a+b*x),data=d,start= list(a =0.1,b=0.2),trace=T); 374 : 0.1 0.2 361 : 0.1 0.2 350 : 0.1 0.2 Error in nls(y ~ rbinom(1000, 1, a + b * x), data = d, start = list(a = 0.1, : step factor 0.000488281 reduced below 'minFactor' of 0.000976562 I have tried plinear, got the same.... Additionaly- why does the parameters not change with the iteration ? [[alternative HTML version deleted]]

peter dalgaard

2012-Jan-03 08:11 UTC

### [R] nls and rbinom function: step factor 0.000488281 reduced below 'minFactor' of 0.000976562

On Jan 3, 2012, at 05:25 , G Vishwanath wrote:> I am trying to learn nls using a simple simulation. I assumed that the binomial prob varies linearly as 0.2 + 0.3*x in x {0,1}, > and the objective is to recover the known parameters a=0.2, b=0.3 > > ..data frame d has 1000 rows... > > d$x<-runif(0,1)...(1000,0,1), I presume.> > > d$y<-rbinom(1000,1,0.2+0.3*d$x) > > table(d$y,cut(d$x,breaks=5)); > > (-0.000585,0.199] (0.199,0.399] (0.399,0.599] (0.599,0.799] (0.799,0.999] > 0 154 149 130 122 114 > 1 34 48 71 76 102 > > z <- nls(y ~ rbinom(1000,1,a+b*x),data=d,start= list(a =0.1,b=0.2),trace=T); >This makes no sense. Random numbers in a model specification???? y~a+b*x might give a result, but you're fitting a model which assumes Gaussian errors with constant variance to data that are nothing of the sort. -- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd.mes at cbs.dk Priv: PDalgd at gmail.com