model fitting result incorrect?

[danatlin]

Hi,
I have successfully fitted the models with runImpulseDE2().
However, upon examining the output, some up-regulated genes were incorrectly classified as down-regulated, and vice versa.
Below are some examples

Examining the fitting results
This one is predicted as up-regulated and reflects the normalized counts.

get_lsModelFits(obj=de.df)$case$Gm2824
$lsImpulseFit
$lsImpulseFit$vecImpulseParam
        beta           h0           h1           h2           t1           t2 
1.018975e+01 2.700061e+01 1.206801e+02 5.721416e-05 2.278622e+01 2.788867e+01 

$lsImpulseFit$vecImpulseValue
        1a         2a         3a        2b        1b        3b    1c    2c    3c 
2.700061e+01 2.700061e+01 2.700061e+01 1.206660e+02 1.206660e+02 1.206660e+02 5.721416e-05 5.721416e-05 5.721416e-05 

$lsImpulseFit$lsvecBatchFactors
NULL

$lsImpulseFit$scaDispParam
  Gm2824 
24.88971 

$lsImpulseFit$scaLL
[1] -22.44949

$lsImpulseFit$scaConvergence
[1] 0

This one predicts up-regulated, but the normalized counts say otherwise.

get_lsModelFits(obj=de.df)$case$Jak3  
$lsImpulseFit
$lsImpulseFit$vecImpulseParam
        beta           h0           h1           h2           t1           t2 
1.167168e+01 6.676505e+00 4.295190e-05 1.500012e+01 1.955861e+01 3.398591e+01 

$lsImpulseFit$vecImpulseValue
         1a         2a         3a        2b        1b        3b    1c    2c    3c 
6.666680e+00 6.666680e+00 6.666680e+00 4.295190e-05 4.295190e-05 4.295190e-05 1.500012e+01 1.500012e+01 1.500012e+01 

$lsImpulseFit$lsvecBatchFactors
NULL

$lsImpulseFit$scaDispParam
    Jak3 
7.301456 

$lsImpulseFit$scaLL
[1] -15.34086

$lsImpulseFit$scaConvergence
[1] 0

Maybe I misunderstood model fitting output?

Thanks,
D