Think, "Dependent variable is OUTCOME variable."
You need to get your terminology correct for discussion; and if you
have it wrong, you might not get the computer programs to run right.
In ANOVA or a general linear model, you want to test the factor, Treatment,
while controlling for the other factors (age, sex, smoking). Some people
like to say "covariate" when the factor is continuous. A dichotomy (also)
can be regarded (and treated) as continuous.
The big error that can be made is to use an analysis that takes the test on
Treatment variance before controlling for the rest. Otherwise, the problems
that are most likely are ones that can arise when Treatment groups are not
generally matched on covariates that affect the outcome. So, for general
information, you should test for differences between Treatments on age,
sex, and smoking -- hopefully, you can make assurances that the groups match.
--
Rich Ulrich
> Date: Sat, 28 May 2016 12:53:32 -0700
> From:
[hidden email]> Subject: statistical testing with multiple covariates
> To:
[hidden email]>
> Hello. I am having some confusion with conducting the statistical test. So,
> just to be clear i am using SPSS 22 and i am supposed to examine whether
> there is any association between the treatment group(intervention and
> control) and the primary outcome, systolic blood pressure(continuous
> variable), with age, sex, smoker status and baseline systolic blood pressure
> as covariates. I would also like to mention that in this case the age is a
> nominal variable ( like <20,20-25 etc) though it can be taken as a
> continuous but i am sticking with the nominal. Though i have used
> independent t-test to determine association between the treatment group and
> systolic blood pressure (with testing blood pressure as testing variable)
> and ANOVA where factor is systolic blood pressure and others are on
> dependent list, i feel this is not the appropriate test to use and maybe i
> should use general linear model for covariates. If anyone could confirm that
> i am doing this the right way it would be great. Thanks
>
>
=====================
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
Locked
