hypothesis testing

hypothesis testing

1. spss always reports 2 tailed

   • 1 tail=direction
     • for 1 tailed divide p(sig) by 2
   • 2 tail= effect

2. read question to evaluate deisgn measure

   • repeated measures design
     • also known as
       • paired samples design
       • within subjects groups
       • dependant groups
     • defining characteristics
       • get all level of IV
       • same people mesaured twice/ using self as control
     • Degrees of freedom = n-1
     • dont caluculate but intreprit in exam
     • measuring same group at least twice
       • looks at sampling distribution of mean diffrence
         • is the diffrence from measurement 1 & 2 diferent to 0
     • effect size
       • looking for consitent change in mean diffrence/ how far apart
       • DO NOT USE COHENS D
     • analysis -->compare means-->paired samples t test
     • assumptions
       • normality
     • µ0 = Null
   • independant group design
     • observational
       • diffrences prexist ie gender, age,birth order
         • no control over group membership
           • quasi experimental
             • differntial
     • experimental
       • random assignment to conditions ie drug vs palcebo
         • researcher controls group membership
           • true experiment
             • aka randomised
     • effect size
       • Cohens D .02 .05 .08
         • how big is the diffrence between groups relative to the SD of either/both groups
           • x̅1 - x̅2 /by σ = D
     • assumptions
       • norrmailty
         • screen data/boxplots, histograms
         • SKEW is the skew statsic 2 times its standard error
       • homogeneity of varience
         • lavenes test = F
           • if levenes is significant the assumption is not met/ REJECT the null (<,.05) (use bottom line of output)
           • non significant means homogentity of varience is met >.05 = allowed to procede with t test
       • Degrees of freedom
         • 1 per group so N-2
     • defining charateristics
       • Analyze --> Compare Means --> Independent-samples T test.
         • dont really do z statistics for IGD
       • 1 level of IV They get the placebo or they dont
         • some ivs cant me manipulated for expirement ie gender age occupation etc
     • NUll = µ1=µ2 or µ1 - µ2 = 0
       • variation from the mean within each group is called error (within group error)
     • aka between groups design
       • compairing 2 groups agaisnt each other

3. step 1

   • state hypothesse (null and alt)
   • step 2
     • step 3
       • step 4
         • evaluate probabilty of null & state conclusion
           • compare to alpha & state conclusion about null & alt hypothese
       • determine probability of null if true
         • take test statistic & work out chance of getting that if the null is true
     • calculate on appropriate test statisic
       • z or t