eric: In the cingulotomy research, which of the following would be important to know to assess the
effectiveness of an MRI-guided cingulotomy for treating psychiatric disease?i) The fraction of psychiatric
patients in general who recover from severe psychiatric symptoms that last a year or longer. ii) The effect
of a cingulotomy on people without mental illness.iii) The number of patients whose follow-up
evaluations were based on letters from them, rather than assessments by a psychiatrist.
Megan Goldman: What do you think on this one?
eric: i realized that this was in one of the previous hw too
eric: so iget why i) is true
eric: and ii)
eric: but not iii), lol
Megan Goldman: Right, that'll help sort out if this treatment has any different effect from whatever most
psych patients are doing.
eric: iii seems somehow irrelevant to me?
Megan Goldman: Well, the follow-up from letters isn't as reliable as from a psychiatrist.
Megan Goldman: If it turns out that a huge portion of the follow-up was done with letters, it may call teh
results into question.
Megan Goldman: Patients may tend to over/understate their symptoms when self-reporting.
eric: oh i see
eric: is this correct if i call this response bias?
Megan Goldman: Self-reporting is prone to response bias, yes.
eric: right ,okay
eric: cool
Sean: Hi Megan
Sean: can I ask you about question 20 on Final 3?
Megan Goldman: Sure.
Sean: it's the which of the "following is closest ..." one
Sean: I think I'm supposd to use the inequalities here, but I
Sean: 'm kinda confused
Sean: i tried doing 16292.64/3992.655

Sean: to get ~4
Sean: then 1/4^2
Sean: and got abou 6.25%
Sean: am i supposed to subtract that from 100% to get the right answer?
Sean: I was using chebychev's inequality
Megan Goldman: Let me noodle it out real quick.
Megan Goldman: you've almost got Chebyshev right... it measures the distance from the mean, so you'll
want (16292.64 - 3915.413)/3992.655 to get a value for k.
Megan Goldman: Then do 1/k^2 to find a probability
....
Recall that the probability is the chance of being
*outside* that many SDs from the mean.
Sean: oh so there is ~10% chance that it's that many SDs outside from the mean
Megan Goldman: So yes, you'll want 1 - that probability to find the chance you're less than that many SDs
away.
Megan Goldman: Right.
Sean: icic
Sean: for the 16292.64-3915 / 3992.655 part
Sean: is that converting to SU?
Megan Goldman: Basically, yes.
Sean: i see
Sean: thank you :)
Megan Goldman: No prob.
Tina: Hi megan
Megan Goldman: Howdy
Tina: can you help me right now or are you with another student?
Megan Goldman: No, go ahead.
Tina: oh ok well I'm having trouble on Q6
Tina: finding the P-value
Tina: I know how to find Z