R help - Jan 2012 - Vegan(ordistep) error: Error in if (aod[1, 5] <= Pin) { : missing value where TRUE/FALSE needed

I am getting the following erro rmessage in ordistep.  I have a number of
similarly structured datasets using ordistep in a loop, and the message
only occurs for some of the datasets.

I cannot include a reproducible sample  - the specific datasets where this
is occur ing are fairly large and there are several pcnm's in the rhs of
the formula.

thanks for any pointers that may allow me to track down the cause of the
error.


Error in if (aod[1, 5] <= Pin) { : missing value where TRUE/FALSE needed
In addition: There were 50 or more warnings (use warnings() to see the
first 50)
traceback()
9: ordistep(myrda0, scope = formula(myrda1), direction = "both",
      Pin = 0.05, Pout = 0.1) at
RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
8: eval(expr, envir, enclos) at RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
7: eval(expr, pf) at RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
6: withVisible(eval(expr, pf)) at
RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
5: evalVis(expr) at RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
4: capture.output(ordistep(myrda0, scope = formula(myrda1), direction
"both",
      Pin = 0.05, Pout = 0.1)) at
RDAPARTIALSexandAgeConnectandGEOGraphy2.R#86
3: eval.with.vis(expr, envir, enclos)
2: eval.with.vis(ei, envir)
1:
source("~/Documents/Dropbox/thesis/CH3/Analysis/RDAPARTIALSexandAgeConnectandGEOGraphy2.R")




print(myrda1)
Call: rda(formula = mygenind@tab ~ pcnmTRE_25_100_CS25 + pcnmTRE_25_10_CS25
+ pcnmTRE_25_2_CS25 +
pcnmTRE_25_5_CS25 + mydata$TreeCov + mydata$Hab_Config +
pcnmEYR_EO_100_CS25 +
pcnmEYR_EO_5000_CS25 + pcnmEYR_TH_10_CS25 + pcnmEYR_TH_2_CS25 +
mydata$Site_No + mydata$Landscape
+ Condition(pcnmCS_NULL + mydata$LAT.x + mydata$LONG.x), na.action
"na.omit")

             Inertia Proportion Rank
Total          1.8110     1.0000
Conditional    0.8681     0.4793   32
Constrained    0.0000     0.0000    0
Unconstrained  0.9429     0.5207   29
Inertia is variance
Some constraints were aliased because they were collinear (redundant)

Eigenvalues for unconstrained axes:
   PC1     PC2     PC3     PC4     PC5     PC6     PC7     PC8
0.16008 0.14733 0.12183 0.09054 0.07380 0.06971 0.05578 0.04215
(Showed only 8 of all 29 unconstrained eigenvalues)


-- 
Nevil Amos
Molecular Ecology Research Group
Australian Centre for Biodiversity
Monash University
CLAYTON VIC 3800
Australia

	[[alternative HTML version deleted]]

Nevil Amos <nevil.amos <at> monash.edu> writes:
> 
> I am getting the following erro rmessage in ordistep.  I have a number of
> similarly structured datasets using ordistep in a loop, and the message
> only occurs for some of the datasets.
> 
> I cannot include a reproducible sample  - the specific datasets where this
> is occur ing are fairly large and there are several pcnm's in the rhs
of
> the formula.
>
> Error in if (aod[1, 5] <= Pin) { : missing value where TRUE/FALSE needed
>
Nevil,

It seems to me that the source of the problem appears in this table:
>              Inertia Proportion Rank
> Total          1.8110     1.0000
> Conditional    0.8681     0.4793   32
> Constrained    0.0000     0.0000    0
> Unconstrained  0.9429     0.5207   29
> Inertia is variance
> Some constraints were aliased because they were collinear (redundant)
> 
The key point is that "Constrained" component is completely alaised
(Inertia 0,
Rank 0) and therefore it cannot be analysed in permutation tests. You get the
same error message with this model:

mod <- rda(dune ~  Moisture + Condition(Moisture), dune.env)

and for the same reason. In your case, PCNM's seem to explain everything and
there is nothing left for other variables, and therefore you cannot analyse
them.

Cheers, Jari Oksanen

I can see how to fix this in vegan. All I can do is to handle these cases
smoothly and with comprehensible error messages, though. They cannot be handled
with permutation tests since there is nothing to do if "Constrained"
component
is zeroed.

Cheers, Jari Oksanen

Whilst the constrained proportion was 0 in the example below I am getting the
error in cases where this is not so.  See new output pasted below


[1] "ORDISTEP RDA1 BOTH WAYS"
[1]
"________________________________________________________________________________________________"
[1] "SFW ALL Adult"
[1] mds3dSFW_EO_100_CS25  mds3dSFW_EO_5000_CS25 mds3dSFW_TH_10_CS25  
mds3dSFW_TH_2_CS25
Call: rda(formula = mygenind@tab ~ mds3dTRE_25_100_CS25 + mds3dTRE_25_10_CS25 +
mds3dTRE_25_2_CS25 + mds3dTRE_25_5_CS25 + mydata$TreeCov + mydata$Hab_Config +
mydata$Site_No + mydata$Landscape + mds3dCS_NULL + mydata$LAT.x + mydata$LONG.x,
na.action
= "na.omit")

              Inertia Proportion Rank
Total          4.5099     1.0000     
Constrained    1.2635     0.2802   29
Unconstrained  3.2464     0.7198  142
Inertia is variance 
Some constraints were aliased because they were collinear (redundant)

Eigenvalues for constrained axes:
    RDA1     RDA2     RDA3     RDA4     RDA5     RDA6     RDA7     RDA8     RDA9
RDA10    RDA11
0.173325 0.126065 0.093396 0.086910 0.072219 0.063064 0.062034 0.058022 0.050743
0.045038 0.043567
   RDA12    RDA13    RDA14    RDA15    RDA16    RDA17    RDA18    RDA19    RDA20
RDA21    RDA22
0.039200 0.035363 0.033358 0.029372 0.028224 0.025715 0.024484 0.024007 0.021999
0.020480 0.018853
   RDA23    RDA24    RDA25    RDA26    RDA27    RDA28    RDA29 
0.017277 0.015510 0.015091 0.012845 0.011089 0.009377 0.006886 

Eigenvalues for unconstrained axes:
   PC1    PC2    PC3    PC4    PC5    PC6    PC7    PC8 
0.2037 0.1710 0.1670 0.1450 0.1343 0.1155 0.1120 0.1015 
(Showed only 8 of all 142 unconstrained eigenvalues)

Error in if (aod[1, 5] <= Pin) { : missing value where TRUE/FALSE needed
In addition: There were 50 or more warnings (use warnings() to see the first
50)
> 
> 
> 
> traceback()
6: ordistep(myrda0, scope = formula(myrda1), direction = "both", 
       Pin = 0.05, Pout = 0.1) at #55
5: eval(expr, envir, enclos) at #55
4: eval(expr, pf) at #55
3: withVisible(eval(expr, pf)) at #55
2: evalVis(expr) at #55
1: capture.output(ordistep(myrda0, scope = formula(myrda1), direction =
"both",
       Pin = 0.05, Pout = 0.1)) at #55
Cheers

Nevil Amos
> 
> Nevil Amos <nevil.amos <at> monash.edu> writes:
> 
>> 
>> I am getting the following erro rmessage in ordistep.  I have a number
of
>> similarly structured datasets using ordistep in a loop, and the message
>> only occurs for some of the datasets.
>> 
>> I cannot include a reproducible sample  - the specific datasets where
this
>> is occur ing are fairly large and there are several pcnm's in the
rhs of
>> the formula.
>> 
>> Error in if (aod[1, 5] <= Pin) { : missing value where TRUE/FALSE
needed
>> 
> Nevil,
> 
> It seems to me that the source of the problem appears in this table:
> 
>>             Inertia Proportion Rank
>> Total          1.8110     1.0000
>> Conditional    0.8681     0.4793   32
>> Constrained    0.0000     0.0000    0
>> Unconstrained  0.9429     0.5207   29
>> Inertia is variance
>> Some constraints were aliased because they were collinear (redundant)
>> 
> 
> The key point is that "Constrained" component is completely
alaised (Inertia 0,
> Rank 0) and therefore it cannot be analysed in permutation tests. You get
the
> same error message with this model:
> 
> mod <- rda(dune ~  Moisture + Condition(Moisture), dune.env)
> 
> and for the same reason. In your case, PCNM's seem to explain
everything and
> there is nothing left for other variables, and therefore you cannot analyse
them.
> 
> Cheers, Jari Oksanen
> 
> I can see how to fix this in vegan. All I can do is to handle these cases
> smoothly and with comprehensible error messages, though. They cannot be
handled
> with permutation tests since there is nothing to do if
"Constrained" component
> is zeroed.
> 
> Cheers, Jari Oksanen
	[[alternative HTML version deleted]]