error cannot run bootsteapping

[chuahmh]

anyone can help ?

error occur during bootstrapping

" a single matrix occur during estimation of path coeefcient...setting another weight scheme coulod solve the problem "

[jmbecker]

There can be several reasons that such a singular matrix problem occurs during the calculations. Possible reasons could be

1. a variable that has zero variance, i.e. it contains only the same values for each case/observation. This can happen, for example, if one uses a grouping variable in the path model and estimates the group models.
2. extreme collinearity, for example, if two variables are perfectly collinear (i.e., they have a correlation of 1).
3. a sample size that is too small. If there are fewer observations than degrees of freedom the matrix of moments is not invertable.

As your problem occurs during the bootrapping it seems that due to the random draw of observations for each resample, problem 1 or problem 2 might be an issue although it is not a problem on the overall sample.
You should also check if there is a variable with relatively low variation or high multicollinearity in you overall model. A too small sample size, however, could also be a problem as it inflates the other problems.

[timbauk]

I encounter the same problem when I bootstrapped my moderating variables (8 variables), although the system run smoothly on my MM and SPM. I use the older version PLS 2.0. Any suggestions how should I handle my data. Need help from the pros'. Thanks.

[timbauk]

The results of my IPMA tend to be between -10 and 90. The total effect analysis provide 2 negative results out of 9 exogenous variables. How could i transform the data to comply the suggested value of 0 to 100. Need help pros. Thanks

[VSVSVS27]

How should I group if the single matrix problem occur???
It is not possible to conduct the MGA,because if I group according to Users (n=133) =1 and Nonusers (n=69)=2, the single matrix problem occurs. How shall I then code the groups ?

[cringle]

A single matrix problem occurs if a variable has identical values (per group). Also, bootstrapping, if a group has a variable that has very similar responses (e.g., 80% a 4 on a 1 to 5 scale), it most likely will happen that bootstrapping randomly draws a subsample that only has identical values for that variable.

It's a data problem, your responses are very similar (per group) for certain variables.

Best
CR

[VSVSVS27]

Thank you, so that means that for a dummy coded variable (user=1, non-users=0) it is not possible to use the MGA?
I do not ask anything on any scale, I understand what you meant by that. But I only grouped them which obviously brings along the same values of either 1 or 0

[jmbecker]

You cannot include the dummy variable that you use for grouping into your model. You either analyse the effect of the grouping dummy variable within your model (you may also use moderating effects of the dummy variable) OR you split your data according to the dummy variable and analyse the effect of all your other variables onto each other (without the dummy in your model), for the different groups.