Page 1 of 2

MLpowsim problem for continuous x random slope model

Posted: Sun Aug 05, 2018 4:49 pm
by guobl
hi Chris and Bill,
I have a problem with MLpowsim for a two level random slope model. the model is similar to the illustration of page 62 the MLpowsim userbook. There is only one continuous x included in the model (random slope). MLpowsim is working with binary X with random slope, but for continuous x (mean= 0, with 0.9= level 1 variance , 0.1= level 2 var), the code ran for 2 seconds and stopped with following message :

Numeric error(s) in calculate command. Affected entries replaced with system missing.

error while obeying batch file analyse.txt at line number 51:
CALC b203=0.600000/b203
Numeric error(s) in calculate command. Affected entries replaced with system missing.
error while obeying batch file analyse.txt at line number 53:
CALC b206=0.600000/b206
Numeric error(s) in calculate command. Affected entries replaced with system missing.
error while obeying batch file analyse.txt at line number 55:
CALC b207=0.600000/b207
Numeric error(s) in calculate command. Affected entries replaced with system missing.
->WSET
if I run a radome intercept only model with continuous X included as fixed effect explanatory, MLpowsim worked very well. just do not know why random slope continuous X not allowed? any help please? I only replace the binary gender with a continuous x in the illustration. my window is window 10 and MLpowsim is the latest version.

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 8:59 am
by ChrisCharlton
Could you please let us know the full set of inputs used to generate the files from MLPowSim, so that we can accurately replicate the problem?

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 9:40 am
by guobl
Thanks Chris, please see my filed uploaded through the system, have to do it one by one as uploading multiple files not allowed. Boliang

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 9:43 am
by guobl
here is the setup file for you Chris

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 9:43 am
by guobl
her are the simu files for you

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 9:44 am
by guobl
here are the graphs file for you chris

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 9:44 am
by ChrisCharlton
Thanks for this. I don't suppose that you could also provide the answers that you initially gave to MLPowSim, as that would allow us to generate the macro files (and regenerate them if we find a bug that needs fixing)?

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 10:45 am
by guobl
Hi Chris,
thanks for looking my query. I am afraid I am not clear what you mean for "answers intitially for MLpowsim" . I originally bet you could find the parameters information from the code. anyway, here are the 2-level model parameter I am using. with numbers in parenthesis are the estimates' values to feed MLpowsim.
y=B0j+B1j*Xij

B0j=B0(0.1)+u0+e
B1j=B1(0.6)+u1


U0 Sgu0(0.08)
U1 cov(0.25) Sgu1(0.02)

Sge=0.8

X~N(0, Var_lev1(1), var_lev2(0.1)) as per Userbook suggestd mean(X)=0

the number of level 2 unit is about 30 to 50 with steps=5
the number of level 1 unit is about 25 to 40 with steps=5
I tried many various number for level 1 and 2 unit numbers.
please let me know any other information you need, thanks,
Boliang

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 11:20 am
by ChrisCharlton
I was hoping for something like:

Code: Select all

 --------------------------------------------------------------------------
|                 MLPowSim version 1.0 Beta 2                              |
|                                                                          |
| coded by William J. Browne and Mousa Golalizadeh (c) February 2018       |
|                                                                          |
| MLPowSim is free software and comes with absolutely NO WARRANTY          |
| MLPowSim produces output files that can be used by the R or MLwiN        |
| packages.                                                                |
| We make no guarantees that the files produced are in any sense correct   |
| or will run in these packages.                                           |
| The further use of any files generated by MLPowSim is the responsibility |
| of the user for whatever purposes they may be used.                      |
 --------------------------------------------------------------------------

To continue using this program having read and understood this
disclaimer please input 1 : 1




                                  Welcome to MLPowSim

Please input 0 to generate R code and 1 to generate MLwiN macros: 1

    Please choose model type

1. 1-level model
2. 2-level balanced data nested model
3. 2-level unbalanced data nested model
4. 3-level balanced data nested model
5. 3-level unbalanced data nested model
6. 3-classification balanced cross-classified model
7. 3-classification unbalanced cross-classified model

Model type : 2
Please input the random number seed: 1
Please input the significance level for testing the parameters: 0.025
Please input number of simulations per setting: 1000




                            Model setup

Please input response type [0 - Normal, 1- Bernouilli, 2- Poisson] : 0
Please enter estimation method [0 - RIGLS, 1 - IGLS, 2 - MCMC] : 1
Do you want to include the fixed intercept in your model (1=YES  0=NO )? 1
Do you want to have a random intercept in your model (1=YES  0=NO )? 1
Do you want to include any explanatory variables in your model (1=YES  0=NO)? 1
How many  explanatory variables do you want to include in your  model?  1
Please choose a type for the predictor x1 (1=Binary  2=Continuous  3=all MVN): 2

Assuming normality, please input its parameters here:

                 The mean of the predictor x1: 0
                 The variance of the predictor x1 at level 1: 0.9
                 The variance of the predictor x1 at level 2: 0.1
Do you want the coefficient associated with explanatory variable x1 to be random (1=YES  0=NO) ? 1




                          Sample size set up

Please input the smallest number of units for the second level: 30
Please input the largest number of units for the second level: 50
Please input the step size for the second level: 5
Please input the smallest number of units for the first level per second level: 25
Please input the largest number of units for the first level  per second level: 40
Please input the step size for the first level  per second level: 5



                             Parameter estimates

Please input estimate of beta_0: 0.1
Please input estimate of beta_1: 0.6


 There is more than one random effect in your model and so you need to enter variance/covariance matrix.
Please input lower triangular entries ( 3 elements):
entry (1,1) is : 0.08
entry (2,1) is : 0.25
entry (2,2) is : 0.02
Please input estimate of sigma^2_e: 0.8




Files to perform power analysis for the 2 level nested model with the following
sample criterion have been created
Sample size in the first level starts at 25 and finishes at 40 with the step size  5

Sample size in the second level starts at 30 and finishes at 50 with the step size  5

1000 simulations for each sample size combination will be performed
Could you confirm that this matches your inputs?

Re: MLpowsim problem for continuous x random slope model

Posted: Mon Aug 06, 2018 12:21 pm
by guobl
yes Chris, that is what I did in the black window!!
by the way, how could you copy these from the black window? i tried many times to copy the procedural information but not success at all!!