Download - R
Transcript
The advice would be to set firstg to 0.02 and expect the necessity to do a second
estimation using the prevAns parameter in siena07. If the problem still occurs for
firstg=0.02, use a smaller value (but less than 0.001 probably makes no sense).
firstg determines the step sizes in the stochastic approximation algorithm; it
is mentioned in some places earlier in this manual. Especially for models with
additional rate effects the default value of 0.2 might be too large. firstg is the
initial value of parameter aN mentioned on p. 393 of Snijders (2001).
Error in solve.default(z$dfra) :
system is computationally singular: reciprocal condition number = 2.34809e–35
(or some other very small number – note that ‘e–35 ’ means 10 to the power –35.)
This can happen at the end of estimation in function siena07, when the covariance
matrix is singular. It means that some effects in the model are linearly related, or are
always 0.
Solutions.
Check your data (look at the description of the variables given as the output of
print01Report). Check your model specification.
14.2
As result of a score-type test (including time test)
Error in solve.default(v9) : Lapack routine dgesv: system is exactly singular
Error in if (cvalue < 0) cvalue <− 0 : missing value where TRUE/FALSE needed.
This can happen as the result of a score test requested in function siena07; or the score
test requested in function sienaTimeTest. In the first case it indicates that there are
linear dependencies in the list of effects (estimated and fixed) that are used for siena07.
If this error message occurs for sienaTimeTest, it indicates linear dependencies in the
list of effects estimated in the siena07 run analyzed by this sienaTimeTest, together
with the interactions with time dummies tested by sienaTimeTest. See Sections 5.9
and 8.2.
Solutions.
For siena07: drop some of the requested score-type test, retaining only a set of tested
effects between which there are no linear dependencies.
For sienaTimeTest: exclude some of the requested time heterogeneity tests by the
excludeEffects parameter.
14.3
In sienaGOF
Error in if (attr(obsData[[groupName]]$depvars[[varName]], ”sparse”)) { :
argument is of length zero
This can happen directly when calling sienaGOF. It indicates that you used a wrong
name for groupName or varName. See the help file for sienaGOF.
Solutions.
Use a correct groupName and varName.
157