- Joined
- Jun 22, 2004
- Messages
- 90
- Reaction score
- 2
Can someone explain this to me. First aid states that if it includes 0, you should accept the null hypothesis. However, in my biostats class, we learned 1.........help!
Question about Confidence Intervals
- Thread starter betz
- Start date
It just depends on how you set the statistics up.
Say we are looking at the risk of having an MI after treatment with a new anti cholesterol drug. Our data says the Risk is 0.34. Awesome.
But the CI is .12 - 1.04. Uh-oh. Here our confidence interval overlaps 1, so the data isn't tha great 🙁 Null Hypothesis time. In this scale, one could never really hit 0
Alternatively, you could set up data where 0 was midline. Say we asked about the same question, and put our data up as the Relative increase or decrease in risk with 0 being normal. Here we might say our mean was -0.45, but the CI was from -2.00 to +0.45. Uh-oh Again! this one overlaps zero, meaning the drug could either raise or lower your risk... so null is right again.
You're just going to have to see how the stats are set up in each question you do 🙂
Say we are looking at the risk of having an MI after treatment with a new anti cholesterol drug. Our data says the Risk is 0.34. Awesome.
But the CI is .12 - 1.04. Uh-oh. Here our confidence interval overlaps 1, so the data isn't tha great 🙁 Null Hypothesis time. In this scale, one could never really hit 0
Alternatively, you could set up data where 0 was midline. Say we asked about the same question, and put our data up as the Relative increase or decrease in risk with 0 being normal. Here we might say our mean was -0.45, but the CI was from -2.00 to +0.45. Uh-oh Again! this one overlaps zero, meaning the drug could either raise or lower your risk... so null is right again.
You're just going to have to see how the stats are set up in each question you do 🙂
- Joined
- Jun 22, 2004
- Messages
- 90
- Reaction score
- 2
Thank you so much! That makes total sense. I just did a qbank question, and they had the data set up where confidence interval overlapped 1. Soooo, I wonder how they would set up data on the actual test. At any rate, now it all makes sense. Thanks!
