Hi
After many reading about logistic model in order to understand them, i have read all and its opposite.
so I post here my results and my own understanding and some additional questions :-).
the IV are:
GDS : a continuous centered variable
MMSE: a continous centered variable.
intc2 : a centered continous variable
the DV is the hit rate (0/1, 1 = success)
here are the results in odds-ratio.
my understanding
when GDS is fixed at its mean, MMSE and intc2 are significant. the interaction is not and diddn't need to be account for. So, I need to only consider the main effect.
results for MMSE means that for one unit increase in the MMSE score, we expect to see about 17% increase in the odds of correct response. right?
I would like to plot the hit response based on the MMSE scores. I have send the following command for different centered score of this IV.
this give the following results
I suppose that the margins are the predictive scores of hit rate based on MMSE.
I would like to be able to compute these scores manually to understand the formula. and I don't knwo how to do that....
hit =b0(25.31)+bMMSEc(1.17)*MMSEc value.
if I take -10.15 for example, I obtain 25.31+(1.17*-10.15) = 25.31+(-11.87) = 13.44. but here, its a odds of hit rate. how could I obtain the mean estimates??
and then I graph the results with the marginsplot directly after the margins command and obtain the joined pdf image.
i'm annoyed because I thought that a logistic model lead to a logistic curve and not something that looks like linear...
I'm clearly not confident in what I'm doing so any feedback should be very helpful.
thanks
carole
After many reading about logistic model in order to understand them, i have read all and its opposite.
so I post here my results and my own understanding and some additional questions :-).
the IV are:
GDS : a continuous centered variable
MMSE: a continous centered variable.
intc2 : a centered continous variable
the DV is the hit rate (0/1, 1 = success)
here are the results in odds-ratio.
Code:
--------------------------------------------------------------------------------
Hit | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
GDSc | .7715028 .1710535 -1.17 0.242 .4995898 1.19141
MMSEc | 1.170367 .0879215 2.09 0.036 1.01013 1.356022
intc2 | 4.220002 1.149595 5.29 0.000 2.474183 7.197694
|
c.MMSEc#c.intc2 | 1.014188 .0370454 0.39 0.700 .9441181 1.089457
|
_cons | 25.30625 11.30362 7.23 0.000 10.54433 60.73465
----------------+----------------------------------------------------------------
_all>N |
var(_cons)| 1.981184 .7668465 .9278009 4.230531
----------------+----------------------------------------------------------------
STIM |
var(_cons)| .6211327 .3838869 .1849715 2.085758
---------------------------------------------------------------------------------
when GDS is fixed at its mean, MMSE and intc2 are significant. the interaction is not and diddn't need to be account for. So, I need to only consider the main effect.
results for MMSE means that for one unit increase in the MMSE score, we expect to see about 17% increase in the odds of correct response. right?
I would like to plot the hit response based on the MMSE scores. I have send the following command for different centered score of this IV.
Code:
margins, at(MMSEc=(-10.15 -7.15 -3.15 0 1.84 3.84))
Code:
-----------------------------------------------------------------------------
| Delta-method
| Margin Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_at |
1 | .7004696 .0862703 8.12 0.000 .5313828 .8695564
2 | .7483582 .0615649 12.16 0.000 .6276932 .8690233
3 | .804184 .0388536 20.70 0.000 .7280323 .8803357
4 | .841726 .0312007 26.98 0.000 .7805738 .9028783
5 | .861123 .0306983 28.05 0.000 .8009555 .9212905
6 | .8801769 .0321054 27.42 0.000 .8172514 .9431024
I would like to be able to compute these scores manually to understand the formula. and I don't knwo how to do that....
hit =b0(25.31)+bMMSEc(1.17)*MMSEc value.
if I take -10.15 for example, I obtain 25.31+(1.17*-10.15) = 25.31+(-11.87) = 13.44. but here, its a odds of hit rate. how could I obtain the mean estimates??
and then I graph the results with the marginsplot directly after the margins command and obtain the joined pdf image.
i'm annoyed because I thought that a logistic model lead to a logistic curve and not something that looks like linear...
I'm clearly not confident in what I'm doing so any feedback should be very helpful.
thanks
carole
Comment