dacohen/Rsquared

module Rsquared
##
  # Tests for deviation of sample mean from expected mean
  # ttest = Rsquared::TTest.new(data, mu0, sided)
  # mu0 is the expected value of the sample mean
  # Supply Rsquared::Upper.tail, Rsquared::Lower.tail or Rsquared::Two.sided
  # Use Upper.tail when you suspect that the sample mean will be greater than the expected mean
  # Use Lower.tail when you suspect that the sample mean will be smaller than the expected mean
  # Use Two.sided when you suspect neither

  class TTest < StatTest
      ##
    # Initializes the TTest object with the supplied arguments
    #

      def initialize(data, mu0, sided)
        @data = data
          @mu0 = mu0
        @sided = sided
        
        if KSTest.new(@data).significant? and @data.length < 40 then
           raise AssumptionError, "The data is not close enough to a normal distribution for such a small sample size"
        end
        if GrubbsTest.new(@data).outlier? then
           raise AssumptionError, "Your data has one or more outliers, which the T-Distribution cannot handle"
        end

        @tstat = (@data.mean - @mu0)/(data.stddev/Math.sqrt(@data.length))
        @pvalue = Distribution::T::cdf(@tstat, @data.length-1)
        self.setSidedness!(@sided)
     end

     ##
     # Returns the t-statistic
     #

     def statistic
         @tstat
     end

     ## significant?, inspect implemented by inheritance

   end
end