Update to julia 1

[kescobo]

This is mostly straightforward. I haven't fixed up the tests yet, but the only thing not working from functions in the README is in this method: thresholds(alt::NoncentralT) = -alt.ncp, alt.ncp, where NoncentralT distribution no longer has ncp

   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.6.0-DEV.1668 (2020-12-04)
 _/ |\__'_|_|_|\__'_|  |  Commit 0253ebf147 (7 days old master)
|__/                   |

julia> using PowerAnalysis, HypothesisTests

julia> n = 100
100

julia> δ = 0.01
0.01

julia> σs = 1.0
1.0

julia> p = 0.8
0.8

julia> α = 0.05
0.05

julia> power(EqualVarianceTTest, n, δ, σs, α)
0.05056741592646641

julia> effect_size(EqualVarianceTTest, n, p, α)
0.39813813756823213

julia> sample_size(EqualVarianceTTest, δ, σs, p, α)
156978.1705208713

julia> type_1(EqualVarianceTTest, n, δ, σs, α)
0.05

julia> type_2(EqualVarianceTTest, n, δ, σs, α)
0.9494325840735336

julia> type_s(EqualVarianceTTest, n, δ, σs, α)
0.41847831065271157

julia> type_m(EqualVarianceTTest, n, δ, σs, α)
ERROR: type NoncentralT has no field ncp
Stacktrace:
 [1] getproperty(x::Distributions.NoncentralT{Float64}, f::Symbol)
   @ Base ./Base.jl:33
 [2] thresholds(alt::Distributions.NoncentralT{Float64})
   @ PowerAnalysis ~/Repos/PowerAnalysis.jl/src/t_test.jl:50
 [3] type_m(::Type{EqualVarianceTTest}, ns::Int64, δ::Float64, σs::Float64, α::Float64, one_sided::Bool)
   @ PowerAnalysis ~/Repos/PowerAnalysis.jl/src/generics.jl:180
 [4] type_m(::Type{EqualVarianceTTest}, ns::Int64, δ::Float64, σs::Float64, α::Float64)
   @ PowerAnalysis ~/Repos/PowerAnalysis.jl/src/generics.jl:170
 [5] top-level scope
   @ REPL[14]:1

julia> exaggeration_factor(EqualVarianceTTest, n, δ, σs, α)
ERROR: type NoncentralT has no field ncp

The type signature changed from here

immutable NoncentralT <: ContinuousUnivariateDistribution
    df::Float64
    ncp::Float64
    function NoncentralT(d::Real, nc::Real)
    	d >= zero(d) && nc >= zero(nc) || error("df and ncp must be non-negative")
        new(float64(d), float64(nc))
    end
end

to here

struct NoncentralT{T<:Real} <: ContinuousUnivariateDistribution
    ν::T
    λ::T
    NoncentralT{T}(ν::T, λ::T) where {T} = new{T}(ν, λ)
end

So is ncp just the new λ?

Still needs:

• What's ncp?
• fix tests syntax
• Add test and docs bots

I'm happy to keep going here, though if you're not interested in holding onto ownership, maybe it could be migrated to the JuliaStats org?

[kescobo]

This might close #1 also

[kescobo]

All tests pass now, but the last command in the README fails

julia> exaggeration_factor(EqualVarianceTTest, n, δ, σs, α)
full precision may not have been achieved in 'pnt{final}'
full precision may not have been achieved in 'pnt{final}'
ERROR: MethodError: no method matching gauss(::PowerAnalysis.var"#f1#12"{Distributions.NoncentralT{Float64}}, ::Float64, ::Float64; rtol=1.0e-8)
Closest candidates are:
  gauss(::Any, ::Any, ::Real, ::Real; rtol, quad) at /home/kevin/.julia/packages/QuadGK/czbUH/src/weightedgauss.jl:60
  gauss(::Integer, ::Real, ::Real) at /home/kevin/.julia/packages/QuadGK/czbUH/src/gausskronrod.jl:110 got unsupported keyword argument "rtol"
  gauss(::Type{T}, ::Integer, ::Real, ::Real) where T<:AbstractFloat at /home/kevin/.julia/packages/QuadGK/czbUH/src/gausskronrod.jl:111 got unsupported keyword argument "rtol"

[kescobo]

Alright, aside from that issue above, I think this is good to go. There's still plenty of work to make this more idiomatic for modern julia, but this is at least functional.

[johnmyleswhite]

So abstractly I'm happy to merge this and turn over ownership to you, but would like some time to go over everything. Ok to follow up next week? Also happy to set up a quick chat over Zoom -- I find using Github isn't a great forum for communication for something with my schedule in which I can block out 30 minutes for deep discussion, but struggle with async comms.

[kescobo]

Yes, definitely fine to take some time, and yes, I'd be happy to meet up via zoom. No rush at all.

I'm US east coast time and I typically have a lot of open space in the schedule on Tuesday/Thursday, and I can be reached on slack/zulip/discourse or via email linked to my GH account.

[johnmyleswhite]

How about we chat on Tuesday next week at 11 AM Eastern?

[johnmyleswhite]

I'll send an e-mail with more details.

[kescobo]

Just summarizing call before merging so I at least have some reference when I forget about this later.

• John has memory of some additional tests that were written, but can't locate them at the moment. I might receive a code dump at some point that I can integrate
• This code should be functional, but there may be extreme examples where some methods fall down, particularly calls to fzero (eg here) for reasons possibly related to numerical instability.
• John is OK with me taking over ownership / maintenance
• Also OK with moving into JuliaStats org if they agree

If I misunderstood anything @johnmyleswhite , feel free to correct. Thanks again for meeting!