Fisher1922

package Fisher1922

Members list

Type members

Classlikes

A model without an explicit Poisson node, and with a fixed number of k=4 dilution stages. Each stage dilutes to 50% of the previous concentration. For fixed arity models, tuple representation works reasonably well. See other, more general models above.

A model without an explicit Poisson node, and with a fixed number of k=4 dilution stages. Each stage dilutes to 50% of the previous concentration. For fixed arity models, tuple representation works reasonably well. See other, more general models above.

This model was created as the first approximation. It is arguably very simple, but also very ineffective (too few stages for our ρ0)

Attributes

Supertypes
class Object
trait Matchable
class Any
Self type

This is an linlined version of the plated model, in which modularization and reuse do not obscure the main structure. There is nothing new here, Just playing with a more concise look for papers and proposals.

This is an linlined version of the plated model, in which modularization and reuse do not obscure the main structure. There is nothing new here, Just playing with a more concise look for papers and proposals.

Also, I find it stunningly pretty! 😊

Attributes

Supertypes
class Object
trait Matchable
class Any
Self type
object FisherPlated

A model with an arbitrary number of stages using vectors (plates). This is the recommended way to handle this model in probula as the types work optimally. The model is over a tuple (ρ0,πs) of randomvariable ρ0 (a Double), and a sequence of binomial parameters π1 ... πk which are all of the same time, so Seq[Double]..

A model with an arbitrary number of stages using vectors (plates). This is the recommended way to handle this model in probula as the types work optimally. The model is over a tuple (ρ0,πs) of randomvariable ρ0 (a Double), and a sequence of binomial parameters π1 ... πk which are all of the same time, so Seq[Double]..

Attributes

Supertypes
class Object
trait Matchable
class Any
Self type
object FisherTupled

A model with an arbitrary number of stages using recursion (hidden in fold) and tuples. Tuples are definitely a bad idea, but it was worth suffering to appreciate this.

A model with an arbitrary number of stages using recursion (hidden in fold) and tuples. Tuples are definitely a bad idea, but it was worth suffering to appreciate this.

The set of dilution parameters πs are all of the same type, but their number is runtime dependent (K), so Tuple is really a painful type to work with (it is useful where the size is fixed, but types differ - we have uniform type, but the size varies). Consequently, Seq is a better type, and we use it to plate (see the model above). The model below mostly shows how not to do it.

Attributes

Supertypes
class Object
trait Matchable
class Any
Self type
final class run

Attributes

Supertypes
class Object
trait Matchable
class Any

Value members

Concrete methods

def eta(ρ0: Double, i: Int): Double

The predictor line for ith group/plate in the Fisher GLM, i = 1..K

The predictor line for ith group/plate in the Fisher GLM, i = 1..K

Attributes

def grid(REnd: Double): Seq[(Double, Seq[Prob])]
def run: Unit

Prints all the histograms computed above.

Prints all the histograms computed above.

Attributes

Concrete fields

val K: Int

Number of stages in each experiment (how many times we dilute), so the length of the geometric series of dilutions.

Number of stages in each experiment (how many times we dilute), so the length of the geometric series of dilutions.

Attributes

val M: SampleSize

The number of times the assay is performed (classicaly M == 1)

The number of times the assay is performed (classicaly M == 1)

Attributes

val N: Int

The number of plates in each dilution stage of the experiment

The number of plates in each dilution stage of the experiment

Attributes

A prior for synthesizing data (known ρ0, for a parameter-recovery test)

A prior for synthesizing data (known ρ0, for a parameter-recovery test)

Attributes

An unformed prior for inference

An unformed prior for inference

Attributes