Converting Odds Ratios to Percentages

[Java Joe]

Hi,

 

I'm working on a logistic regression model and would appreciate help on converting the odds ratios.  My understanding from class is that an odds ratio of 1.25 can be intepreted as 25 percent.  But I'm not clear on what to do when the odds ratio is more than 1.0, like 4.37 or 3.09.  How do I convert them to percentages?

 

Sorry for such a simple question  but if you can help me understand this better, it'll be a big help.

 

Thank you!

 

Joe(sy)

[Bruce Weaver]

Java Joe wrote

Hi,

I'm working on a logistic regression model and would appreciate help on
converting the odds ratios.  My understanding from class is that an odds
ratio of 1.25 can be intepreted as 25 percent.  But I'm not clear on what to
do when the odds ratio is more than 1.0, like 4.37 or 3.09.  How do I
convert them to percentages?

Sorry for such a simple question  but if you can help me understand this
better, it'll be a big help.

Thank you!

Joe(sy)

How did you get 25% for 1.25?  Presumably, you did this:

   (1.25 - 1.00) x 100 = 25%

So generally:

  (OR - 1) x 100 = percent increase (over the reference category) in the odds of the outcome

p.s. - Last time I checked, 1.25 was greater than 1.  ;-)  

[Java Joe]

Thanks, Bruce!  Your  explanation makes more sense than my class notes.  :)   joesy

On Tue, Mar 9, 2010 at 5:07 AM, Bruce Weaver <[hidden email]> wrote:

Java Joe wrote:
>
> Hi,
>
> I'm working on a logistic regression model and would appreciate help on
> converting the odds ratios.  My understanding from class is that an odds
> ratio of 1.25 can be intepreted as 25 percent.  But I'm not clear on what
> to
> do when the odds ratio is more than 1.0, like 4.37 or 3.09.  How do I
> convert them to percentages?
>
> Sorry for such a simple question  but if you can help me understand this
> better, it'll be a big help.
>
> Thank you!
>
> Joe(sy)
>
>

How did you get 25% for 1.25?  Presumably, you did this:

  (1.25 - 1.00) x 100 = 25%

So generally:

 (OR - 1) x 100 = percent increase (over the reference category) in the
odds of the outcome

p.s. - Last time I checked, 1.25 was greater than 1.  ;-)

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

NOTE:  My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.
--

View this message in context: http://old.nabble.com/Converting-Odds-Ratios-to-Percentages-tp27830618p27836982.html

Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[GiantRolo]

Hi Bruce,

I am having trouble understanding odds ratio's in relation to percentages.

I think I'm right in thinking that an odds ratio of 1.07 corresponds to a 7% increase in the odds of whatever your variable is occurring.

But just so I can be clear on your response to Java Joe, in the examples he quoted where the odds ratio's were higher (4.37 or 3.09), would they then correspond to percentages of 337% or 209%?

Many thanks,
GiantRolo

[Hector Maletta]

GiantRolo,
Yours is the right definition of odds, but I feel you need some additional
precision to correctly understand the issue. First, the odds of an event X
equals the probability of X divided by the probability of non-X. Thus, odds
of X = 1.07 means that the probability of X divided by the probability of
non-X is 1.07. This means that the prob of X is p(X)=1.07 / (1+1.07) =
1.07/2.07=0.5169. Your other examples with odds of 4.37 and 3.09 are correct
as well. Second, the odds are always relative to some reference. For
instance, in a binary variable such as sex, the odds of being male =
p(male)/p(female). In a multinomial variable such as Region (with categories
such as East, North, South and West) you must choose a reference category,
say East. Thus the odds of being West = p(West)/p(East), or more generally,
p(X category)/p(reference category). Third, the odds are always the odds of
a VALUE, or CATEGORY, or EVENT, not the odds of a variable: a variable by
definition has more than one value, and the odds measure the relative
likelihood of one particular value, relative to the likelihood of a
reference value. Fourth, the odds may be computed for a SET of values, as
distinct from a particular value; for instance, the odds of being either
West or South, relative to the reference region (East), are odds(West OR
South) = p(West OR South)/p(East). Assuming regions are exclusive of each
other, i.e. cases belong to one region only, this would equal
=[p(West)+p(South)] / p(East).

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de
GiantRolo
Enviado el: Wednesday, October 31, 2012 14:15
Para: [hidden email]
Asunto: Re: Converting Odds Ratios to Percentages

Hi Bruce,

I am having trouble understanding odds ratio's in relation to percentages.

I think I'm right in thinking that an odds ratio of 1.07 corresponds to a 7%
increase in the odds of whatever your variable is occurring.

But just so I can be clear on your response to Java Joe, in the examples he
quoted where the odds ratio's were higher (4.37 or 3.09), would they then
correspond to percentages of 337% or 209%?

Many thanks,
GiantRolo





--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/Converting-Odds-Ratios-to-Perc
entages-tp1092350p5715977.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command SIGNOFF SPSSX-L For a list of
commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD