topic Re: Comparison of VCA between JMP and R in Discussions

Comparison of VCA between JMP and R

Mickyboy — Fri, 09 Jun 2023 00:31:28 GMT

Hi,

l am looking at a study that was performed in R, l have managed to replicate the results of the VCA in R using the following syntax, VisDil1 <- lmer(l2t ~ (1|Op/Day/MDCK) , Dil1), for the following results

Random effects:

Groups Name Variance Std.Dev.

MDCK:(Day:Op) (Intercept) 0.02309 0.15195

Day:Op (Intercept) 0.00861 0.09279

Op (Intercept) 0.00000 0.00000

when i run what i believe is the same model in JMP i get

Random Effect Var Ratio    Var Component Pct of Total
MDCK[Day,Op]    0.1904762    0.031746 16.000
Day[Op]    0    0    0.000
Op    0    0    0.000
Residual 0.1666667    84.000
Total 0.1984127   100.000

l used the following script

Fit Model(
	Y( :l2t ),
	Effects,
	Random Effects( :MDCK[:Day, :Op], :Day[:Op], :Op ),
	NoBounds( 0 ),
	Estimate Only Variance Components( 1 ),
	Personality( "Standard Least Squares" ),
	Method( "REML" ),
	Run Model
);

i looked at this across 8 dilutions, and it is the same result in each dilution, R has a variance component for Day:Op, whilst jmp is combining this effect into it into the effect MDCK[Day,Op], so l feel i am not specifying the model in JMP the same as in R to get the same result.

Does anyone have any advice???

Thanks

Re: Comparison of VCA between JMP and R

Mickyboy — Thu, 01 Apr 2021 05:33:32 GMT

when i posted it has cut the residual component off, from R, here it is again
Random effects:
Groups Name Variance Std.Dev.
MDCK:(Day:Op) (Intercept) 0.02309 0.15195
Day:Op (Intercept) 0.00861 0.09279
Op (Intercept) 0.00000 0.00000
Residual 0.16670 0.40829

Re: Comparison of VCA between JMP and R

Mark_Bailey — Thu, 01 Apr 2021 17:26:46 GMT

Please clarify the structure of the design and the randomization.

Op is at the top of the hierarchy.

The 4 Days are different for both Op.

The two MDCK are different for each day and Op.

Isn't titer a potentially important fixed effect?

Re: Comparison of VCA between JMP and R

Mickyboy — Tue, 06 Apr 2021 00:33:28 GMT

Hi Mark,

Thanks for your reply, please find attached a copy of the nested design, l agree that the syntax dosent reflect the design of the experiment, l think a more appropriate model would be

RBC
Day[RBC]
Op[RBC,Day]

However, thats another story, what l am trying to understand is why when l run what l believe is the same model in both R and JMP, I get a variance component for Day in R, but JMP seems to combine this component into the MDCK variable.

The variable l2t is the log2 of titer, the independent variable.

Thanks