|
|
Re: Our first control group result is precisely how we want AML's
program to behave with control group data ....
Posted:
Apr 29, 2012 1:58 AM
|
|
On Apr 28, 7:45 pm, djh <halitsk...@att.net> wrote:
> In a post above, you wrote:
>
> "Only if both coefficients are individually significant and they have
> different signs can you skip testing the
> difference."
>
> Using the {lnL,mv,lnLmv} model, the study and control group results
> for the b1 fold appear to exemplify this kind of case, where
> corresponding coefficients are individually significant with flipped
> signs, thereby eliminating the need for further testing.
>
> b1 study group results for {lnL,mv,lnLmv} model
>
> 125831 cases have Y=0; 29715 cases have Y=1.
>
> Overall Model Fit...
> Chi Square=34591.3722; df=3; p= 0.0000
>
> Coefficients and Standard Errors...
> Variable Coeff. StdErr p
> 1 4.3831 0.0460 0.0000
> 2 0.3439 0.0979 0.0004
> 3 -0.0734 0.0234 0.0017
> Intercept -19.3877
>
> Odds Ratios and 95% Confidence Intervals...
> Variable O.R. Low -- High
> 1 80.0878 73.1806 87.6470
> 2 1.4104 1.1643 1.7086
> 3 0.9292 0.8875 0.9729
>
> b1 control group results for {lnL,mv,lnLmv} model
>
> 24854 cases have Y=0; 3237 cases have Y=1.
>
> Overall Model Fit...
> Chi Square= 4256.4807; df=3; p= 0.0000
>
> Coefficients and Standard Errors...
> Variable Coeff. StdErr p
> 1 3.8102 0.1278 0.0000
> 2 -1.3225 0.2610 0.0000
> 3 0.3564 0.0655 0.0000
> Intercept -17.0582
>
> Odds Ratios and 95% Confidence Intervals...
> Variable O.R. Low -- High
> 1 45.1573 35.1488 58.0156
> 2 0.2665 0.1598 0.4445
> 3 1.4282 1.2561 1.6240
>
> Have I interpreted your advice correctly, i.e. is this a case that
> does not require further testing of individual corresponding
> coefficients, inasmuch as all four are significant and corresponding
> pairs have flipped signs?
Yes, this is such a case.
|
|
