I'm kind of new to Stats and R in general and I'm looking to clear some doubts. I'm trying to use Censored regression models to do the analysis of my qPCR data. The response variable is continuous (Ct values), the value that the PCR machine gives. This is the code I'm using:

reg1 <- tobit(PCR ~ 0 + Scenario + Dilution, left = -Inf, right =37 , data = dilution)

My question is; if the threshold for considering a test result Pos/Neg is: Positive < 37 (Ct), Negative >= 37 (Ct), and the PCR machine runs up to 45 Cycles, what should be the limit I set for the censored data?? when I set up the limit for censoring on 37 I get this result:

Error in solve.default(vcov.hyp) : Lapack routine dgesv: system is exactly singular: U[10,10] = 0

And if I set up the limit for censoring on 38 is the same, but when I put the limit for censoring on 39 or higher, then it works. Why is this happening? what does this mean? I think I'm doing something really wrong. I'm looking for answers in the style of "Stats for Dummies" :/

Ah, the PCR results data, the Ct values, sometimes have a value that is 'undetermined', should I substitute those values for the 45 (max number of cycles of the PCR machine)? or should I leave those as N/A? the proble I see is that when I leave those 'undetermined' results as 'undetermined', then the R eliminates those and tells me this:

Observations: (13 observations deleted due to missingness)

The idea of doing censored regression was not to lose data points, at least I thought that in the beginning. I know that the estimates that I would obtain from the tobit model would be better than from a OLS in the case of my data. That's why I want to use it.

Please help. Thanks in advance for the collaboration.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.