########
> str(MC)
'data.frame': 21130 obs. of 10 variables:
$ male : int 1 1 1 1 1 1 1 1 1 1 ...
$ pop : int 2 2 2 2 2 2 2 2 2 2 ...
$ lma : num 4.9 4.9 4.9 4.9 4.9 4.9 4.9 4.9 4.9 4.9 ...
$ poids : num 3.03 3.03 3.03 3.03 3.03 3.03 3.03 3.03 3.03 3.03 ...
$ ventre: int 4 4 4 4 4 4 4 4 4 4 ...
$ dos : int 3 3 3 3 3 3 3 3 3 3 ...
$ date : int 1 1 1 1 1 1 1 1 1 1 ...
$ mom : int 1 1 1 1 1 2 2 2 2 2 ...
$ c : chr "no" "no" "no" "no" ...
$ N : Factor w/ 5 levels "explo","stress",..: 3 1 4 5 2 3 1 4 5 2 ...
########
"your inputs have redundant/overlapping information."
yes, thas the case, as you can see :
> cor.test(MC$lma,MC$poids)
Pearson's product-moment correlation
data: MC$lma and MC$poids
t = 168.5517, df = 21128, p-value < 2.2e-16
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
0.7514854 0.7629881
sample estimates:
cor
0.7572954
########
"If your system is exactly singular, it means that at least one of the
columns of your 'model matrix'á is some linear combination of the others, which
essentially means your inputs have redundant/overlapping information."
I think that my variables have some redundant information but it seems me that
type-II analysis-of-variance permit to bypass this issue.
Are your input variables supposed to be factors (categorical) or numeric?
In the model I show, there is only continues numeric independent variables (lma
and poids). The dependent variable is an categorical and unordered.
########
BUT, when that's will work, I'd like add some independent categorical and
ordered variable (it's possible to considerer them as unordered)
Does it imply some issues ??
########
This can also happen if one of the response categories only occurs with a
single combination of your
inputs. Runá some tables on the variables in your model; that might help
clarify things.
But, lma and poids are continues variables
> table(MC$lma,MC$poids)
1.97 2.17 2.35 2.37 2.41 2.46 2.47 2.48 2.54 2.59 2.71 2.8 2.83 2.85
4 560 0 0 0 0 0 0 0 0 0 0 0 0 0
4.4 0 0 0 0 0 0 0 475 0 0 0 0 0 0
4.5 0 520 0 0 0 0 0 0 0 0 0 0 0 0
4.6 0 0 570 0 560 0 520 0 0 0 0 0 0 475
4.7 0 0 0 0 0 565 0 0 0 0 570 0 0 0
4.8 0 0 0 0 0 0 1090 0 0 570 0 560 0 0
4.9 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5 0 0 0 560 0 0 0 0 0 0 0 0 0 0
5.1 0 0 0 0 0 0 0 0 0 0 0 0 570 0
5.2 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5.4 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5.5 0 0 0 0 0 0 0 0 515 0 0 0 0 0
2.88 2.91 3 3.02 3.03 3.09 3.11 3.15 3.16 3.18 3.2 3.24 3.35 3.4
4 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4.4 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4.6 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4.7 0 0 0 0 0 475 515 0 0 0 0 0 0 0
4.8 570 0 0 0 0 0 0 0 0 0 0 0 0 0
4.9 0 560 515 0 1550 0 0 0 0 0 570 560 0 0
5 0 0 0 520 0 0 0 0 0 0 570 0 560 0
5.1 0 0 0 0 0 0 0 520 0 0 0 0 0 520
5.2 0 0 0 0 0 0 0 0 570 570 0 0 0 0
5.4 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3.43 3.67 3.71 3.95 4 4.06
4 0 0 0 0 0 0
4.4 0 0 0 0 0 0
4.5 0 0 0 0 0 0
4.6 0 0 0 0 0 0
4.7 0 0 0 0 0 0
4.8 0 0 0 0 0 0
4.9 0 0 0 0 0 0
5 0 0 0 0 0 0
5.1 0 0 0 0 0 0
5.2 570 560 0 520 0 0
5.4 0 0 565 0 0 520
5.5 0 0 0 0 570 0
> table(MC$poids,MC$N)
explo stress SS_cache thermo immo
1.97 112 112 112 112 112
2.17 104 104 104 104 104
2.35 114 114 114 114 114
2.37 112 112 112 112 112
2.41 112 112 112 112 112
2.46 113 113 113 113 113
2.47 322 322 322 322 322
2.48 95 95 95 95 95
2.54 103 103 103 103 103
2.59 114 114 114 114 114
2.71 114 114 114 114 114
2.8 112 112 112 112 112
2.83 114 114 114 114 114
2.85 95 95 95 95 95
2.88 114 114 114 114 114
2.91 112 112 112 112 112
3 103 103 103 103 103
3.02 104 104 104 104 104
3.03 310 310 310 310 310
3.09 95 95 95 95 95
3.11 103 103 103 103 103
3.15 104 104 104 104 104
3.16 114 114 114 114 114
3.18 114 114 114 114 114
3.2 228 228 228 228 228
3.24 112 112 112 112 112
3.35 112 112 112 112 112
3.4 104 104 104 104 104
3.43 114 114 114 114 114
3.67 112 112 112 112 112
3.71 113 113 113 113 113
3.95 104 104 104 104 104
4 114 114 114 114 114
4.06 104 104 104 104 104
> table(MC$lma,MC$N)
explo stress SS_cache thermo immo
4 112 112 112 112 112
4.4 95 95 95 95 95
4.5 104 104 104 104 104
4.6 425 425 425 425 425
4.7 425 425 425 425 425
4.8 558 558 558 558 558
4.9 751 751 751 751 751
5 442 442 442 442 442
5.1 322 322 322 322 322
5.2 558 558 558 558 558
5.4 217 217 217 217 217
5.5 217 217 217 217 217
Date: Fri, 5 Feb 2010 05:36:12 -0800
Subject: Re: [R] mlogit
From: [email protected]
To: [email protected]
Hi:
On Fri, Feb 5, 2010 at 5:23 AM, hugo Mathe <[email protected]> wrote:
Hi,
I'm using the function "mlogit" from the package "mlogit" in order to make a
multinomial model, with random and nested effect*. But, currently, even a basic
model as
> mlogit(c ~ lma + poids , MC, shape = "long", alt.var = "N")
Erreur dans drop(.Call("La_dgesv", a, as.matrix(b), tol, PACKAGE = "base")) :
ásous-programme Lapack dgesv : le systÞme est exactement singulier
which is the same thing than
Error in drop(.Call("La_dgesv", a, as.matrix(b), tol, PACKAGE = "base")) :
"Lapack routine dgesv: system is exactly singular"
If your system is exactly singular, it means that at least one of the
columns of your 'model matrix'á is some linear combination of the others, which
essentially means your inputs have redundant/overlapping information. Are
your input variables supposed to be factors (categorical) or numeric? Type
str(MC) to view the class of each of the variables in your data frame to see
whether your expectations match reality. This can also happen if one of
the response categories only occurs with a single combination of your
inputs. Runá some tables on the variables in your model; that might help
clarify things.
HTH,
Dennis
I didn't find any explanations on the web ...
my data look like this
> head(MC)
ámale pop lma poids ventre dos date mom á c á á áN
1 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á1 á á á áno áSS_cache
2 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á1 á á á áno á á á áexplo
3 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á1 á á á áno á á thermo
4 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á1 á á á áno á á á áimmo
5 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á1 á á á áno á á á ástress
6 á á1 á 2 á á4.9 á3.03 á á á4 á á 3 á á á1 á á2 á á á yes á SS_cache
somes data of the Examples (in ?mlogit) look like mine
> head(TravelMode)
áindividual ámode choice wait vcost travel gcost income size avincome á á time
1 á á á á á1 á air á á no á 69 á á59 á á100 á á70 á á3.5 á á1 á á á 35 2.816667
2 á á á á á1 train á á no á 34 á á31 á á372 á á71 á á3.5 á á1 á á á á0 6.766667
3 á á á á á1 á bus á á no á 35 á á25 á á417 á á70 á á3.5 á á1 á á á á0 7.533333
4 á á á á á1 á car á áyes á á0 á á10 á á180 á á30 á á3.5 á á1 á á á á0 3.000000
5 á á á á á2 á air á á no á 64 á á58 á á 68 á á68 á á3.0 á á2 á á á 30 2.200000
6 á á á á á2 train á á no á 44 á á31 á á354 á á84 á á3.0 á á2 á á á á0 6.633333
á timeair
1 2.816667
2 0.000000
3 0.000000
4 0.000000
5 2.200000
6 0.000000
and
> mlogit(choice ~ time + timeair , TravelMode, shape = "long", alt.var = "mode")
Call:
mlogit(formula = choice ~ time + timeair, data = TravelMode, á á shape =
"long", alt.var = "mode", method = "nr", print.level = 0)
Coefficients:
alttrain á áaltbus á áaltcar á á átime á timeair
á-3.6776 á -4.2339 á -4.5232 á -0.5963 á -2.7048
work prefectly
so, I don't understant any thing, ...
Is anybody understand why this function didn't work on my data ?
MANY THANKS
hugo mathÚ hubertEmail : [email protected]
PS: * if anybody know how to consider interaction between 2 nested factors, I'm
interested ...
(it is "a:b in (A:B)" if a is nested in A and b nested in B ?)
_________________________________________________________________
Hotmail: Powerful Free email with security by Microsoft.
á á á á[[alternative HTML version deleted]]
______________________________________________
[email protected] mailing list
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Hotmail: Trusted email with powerful SPAM protection. Sign up now.
_________________________________________________________________
Hotmail: Trusted email with Microsofts powerful SPAM protection.
[[alternative HTML version deleted]]
______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
