tamaan loading model prior and non-parametric IRT (tam.np)

[andreanorcinipala]

Hi,
I am encountering two separate issues: How to specify model priors with tamaan and use tam.np.

RE: model prior. I could not find any example on how to specify lambdas' model prior in tamaan. I tried to come up with potential ways, but it is not working at all.

One example is:
tammodel <- "
F1=~A1__A10
F1 ~~ 1*F1

MODEL PRIOR:
p1arace1_F1_load__p9arace1_F1_load ~ N(0,2);
"

Another example is:
tammodel <- "
F1=~lamb1__lamb10A1__A10
F1 ~~ 1F1

MODEL PRIOR:
lamb1__lamb10 ~ N(0,2);
"
May you provide an example of how to specify lambas' prior?

The second issue is related to tam.np. I have tried to run example 1 (TAM page 177 https://cran.r-project.org/web/packages/TAM/TAM.pdf)

data(data.cqc02, package="TAM")
dat <- data.cqc02

mod <- TAM::tam.np(dat)

But I keep getting an error:
Error: Not compatible with requested type: [type=NULL; target=double].
In addition: Warning message:
In max(sqrt((probs - probs0)^2 * pi_k_array)) :
no non-missing arguments to max; returning -Inf

Can you please help with this?
Thanks
Andrea

[alexanderrobitzsch]

RE: model prior. I could not find any example on how to specify lambdas' model prior in tamaan. I tried to come up with potential ways, but it is not working at all.

One example is:
tammodel <- "
F1=~A1__A10
F1 ~~ 1*F1

MODEL PRIOR:
p1arace1_F1_load__p9arace1_F1_load ~ N(0,2);
"

Another example is:
tammodel <- "
F1=~lamb1__lamb10_A1__A10 F1 ~~ 1_F1

MODEL PRIOR:
lamb1__lamb10 ~ N(0,2);
"
May you provide an example of how to specify lambas' prior?

The issue is that the __ operator is (now) not defined for parameters, only for variables. Hence, lamb1__lamb10 is not understood. One should write separate model statements for each parameter. To avoid copying, one could use the DO operator which is a shortcut in tamaan syntax.

[alexanderrobitzsch]

The second issue is related to tam.np. I have tried to run example 1 (TAM page 177 https://cran.r-project.org/web/packages/TAM/TAM.pdf)

data(data.cqc02, package="TAM")
dat <- data.cqc02

mod <- TAM::tam.np(dat)

But I keep getting an error:
Error: Not compatible with requested type: [type=NULL; target=double].
In addition: Warning message:
In max(sqrt((probs - probs0)^2 * pi_k_array)) :
no non-missing arguments to max; returning -Inf

I agree. This was indeed a bug. The fix is included in recent TAM 3.6-8.

[andreanorcinipala]

RE: model prior. I could not find any example on how to specify lambdas' model prior in tamaan. I tried to come up with potential ways, but it is not working at all.
One example is:
tammodel <- "
F1=~A1__A10
F1 ~~ 1*F1
MODEL PRIOR:
p1arace1_F1_load__p9arace1_F1_load ~ N(0,2);
"
Another example is:
tammodel <- "
F1=~lamb1__lamb10_A1__A10 F1 ~~ 1_F1
MODEL PRIOR:
lamb1__lamb10 ~ N(0,2);
"
May you provide an example of how to specify lambas' prior?

The issue is that the __ operator is (now) not defined for parameters, only for variables. Hence, lamb1__lamb10 is not understood. One should write separate model statements for each parameter. To avoid copying, one could use the DO operator which is a shortcut in tamaan syntax.

Thank you so much for your prompt reply. Unfortunately, even with your suggested syntax I do not seem to make priors lambda work. That is, lambdas remain the same regardless of the priors I use (e.g., ~N(0,2), ~N(0,1), ~N(0,0.1)). Would you be able to provide an example that works? Thanks!