1. Dismiss Notice
  2. Download free Tapatalk for iPhone or Tapatalk for Android for your phone and follow the SDN forums with push notifications.
    Dismiss Notice

is this correct regarding null hypothesis and alpha errors?

Discussion in 'Step I' started by suckerfree, Aug 3, 2006.

  1. suckerfree

    10+ Year Member

    Joined:
    Sep 10, 2004
    Messages:
    160
    Likes Received:
    0
    Hi,

    Could anyone tell me if this is wrong because this is how I think of this confusing concept:

    -------
    Null hypothesis: there is no relationship between risk factor and disease state; the relationship you're trying to establish is not there.

    Type I alpha error: you made a mistake to have rejected the null (error of commision)

    Type II beta error: you made a mistake to have accepted the null (error of ommission)

    If p < .05, you should accept the null
    p = probability of commiting an alpha error; probability of being wrong to reject the null and believe there is a relationship between the risk factor and the disease
     
  2. Note: SDN Members do not see this ad.

  3. superoxide

    superoxide Member
    5+ Year Member

    Joined:
    Jun 29, 2006
    Messages:
    535
    Likes Received:
    0

    You are correct. Here's just another way of looking at it (this is the way I remember it from FA):


    Type 1 error: 'you saw something that didnt exist' .There was no relationship, but you saw one. you should have accepted the null hypothesis, but you rejected it.

    Type 2 error: 'you did not see a difference that did exist' There was a relationship, but you didnt see it. You should have rejected the null hypothesis, but you accepted it.
     
  4. trudub

    trudub Senior Member
    7+ Year Member

    Joined:
    Oct 23, 2003
    Messages:
    337
    Likes Received:
    2
    Not quite right though. You say if p < .05 then ACCEPT the null. In fact, if p < .05 (if that is what you have set as your alpha value) then you should REJECT the null. If p < .05 that means that there is less than a 5% chance that the observed difference is by chance.
     

Share This Page