[R] Issue about eRM package

[Michael Dewey]

I do not know that package so cannot help with that but the dataset did 
not come through. This mailing list is quite restrictive in what sorts 
of attachment it allows so I suggest trying something like a .csv file 
or a .txt file.

Michael

On 28/09/2021 10:37, 宋启发 wrote:
>   I am using your excellent R package eRM to solve a questionaire survey data. I meet a strange issue when using RSM function. This RSM runs well using sample data.
> 
> There is a test data like this:
> 
> 'data.frame':   1669 obs. of  7 variables:
> 
>   $ X1: num  2 2 3 3 2 3 4 4 3 2 ...
> 
>   $ X2: num  4 3 3 2 3 4 4 4 3 3 ...
> 
>   $ X3: num  4 3 3 4 3 3 4 4 3 0 ...
> 
>   $ X4: num  2 3 2 4 1 2 3 3 3 1 ...
> 
>   $ X5: num  4 4 3 3 3 4 4 4 3 3 ...
> 
>   $ X6: num  4 4 4 3 3 4 4 4 3 3 ...
> 
>   $ X7: num  4 2 3 4 3 3 4 4 3 3 ...
> 
>   
> 
> ###
> 
> RSM(test)
> 
> Warning in sqrt(diag(solve(parest$hessian))) : NaNs produced
> 
> Warning in sqrt(diag(lres$W %*% solve(parest$hessian) %*% t(lres$W))) :
> 
>    NaNs produced
> 
>   
> 
> Results of RSM estimation:
> 
>   
> 
> Call:  RSM(X = test)
> 
>   
> 
> Conditional log-likelihood: 13768080
> 
> Number of iterations: 5
> 
> Number of parameters: 9
> 
>   
> 
> Item (Category) Difficulty Parameters (eta):
> 
>                  X2        X3       X4        X5        X6        X7     Cat 2     Cat 3       Cat 4
> 
> Estimate -700.9534 -456.4338 901.4649 -1322.033 -1256.828 -673.2412 -231.5791 -3979.953 1911.748203
> 
> Std.Err        NaN       NaN      NaN       NaN       NaN       NaN       NaN       NaN    1.036215
> 
>   
> 
> ###
> 
> So, I use a shortened data, run like this:
> 
> RSM(test[1:283,])
> 
> Results of RSM estimation:
> 
>   
> 
> Call:  RSM(X = test[1:283, ])
> 
>   
> 
> Conditional log-likelihood: -1015.388
> 
> Number of iterations: 29
> 
> Number of parameters: 9
> 
>   
> 
> Item (Category) Difficulty Parameters (eta):
> 
>                   X2         X3        X4         X5         X6         X7     Cat 2     Cat 3    Cat 4
> 
> Estimate -0.0739008 0.05655997 1.2203061 -1.1212188 -0.7547957 -0.0378593 1.2020940 3.2657449 8.501760
> 
> Std.Err   0.1018267 0.10026033 0.1003035  0.1195827  0.1125677  0.1013739 0.4387487 0.8161754 1.226057
> 
>   
> 
> ###
> 
> However, when I add one record, from 283 to 284,
> 
> RSM(test[1:284,])
> 
> Warning in sqrt(diag(solve(parest$hessian))) : NaNs produced
> 
> Warning in sqrt(diag(lres$W %*% solve(parest$hessian) %*% t(lres$W))) :
> 
>    NaNs produced
> 
>   
> 
> Results of RSM estimation:
> 
>   
> 
> Call:  RSM(X = test[1:284, ])
> 
>   
> 
> Conditional log-likelihood: 3369344
> 
> Number of iterations: 6
> 
> Number of parameters: 9
> 
>   
> 
> Item (Category) Difficulty Parameters (eta):
> 
>                  X2        X3         X4        X5        X6        X7     Cat 2     Cat 3       Cat 4
> 
> Estimate -670.9137 -554.6215 527.789997 -1359.721 -1127.137 -626.1859 -126.7199 -4753.367 2134.408482
> 
> Std.Err  2152.9823 1679.2875   2.377241       NaN       NaN 3394.5180  562.6800 11045.935    3.526018
> 
>   
> 
> I can’t find any special values in the data list
> 
>     X1 X2 X3 X4 X5 X6 X7
> 
> 270  4  4  4  3  4  4  4
> 271  3  3  3  1  3  3  3
> 272  4  3  4  1  4  4  4
> 
> 273  3  3  3  3  3  3  3
> 
> 274  3  3  4  4  3  3  4
> 
> 275  3  3  3  3  3  3  3
> 
> 276  1  3  3  3  3  3  3
> 
> 277  3  3  3  2  4  4  4
> 
> 278  2  3  2  1  3  3  3
> 
> 279  3  3  3  2  3  3  3
> 
> 280  3  3  2  2  3  3  3
> 
> 281  3  4  4  3  4  3  3
> 
> 282  2  2  2  2  2  2  2
> 
> 283  3  4  3  1  4  3  3
> 
> 284  2  4  2  1  4  3  2
> 
> 285  3  3  3  3  3  3  3
> 
> 286  4  4  3  4  4  4  4
> 
> 287  4  3  4  0  3  4  4
> 
> 288  0  3  4  0  4  4  1
> 
> 289  4  4  4  4  4  4  4
> 
> 290  3  3  3  3  3  3  2
> 
>   
> 
> If I input many different numbers of data, the results often become strange.
> 
> The data is appended at the letter. No na values in all data.
> 
>   
> 
> Great thanks
> 
> 
> 
> #### all data  is in the appended file
> 
> 
> 
> 
> 
> 
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
> 
> 

-- 
Michael
http://www.dewey.myzen.co.uk/home.html