Help the Stat Consulting Group by
giving a gift
Loading
|
|
|
||||
|
Help the Stat Consulting Group by
giving a gift
| |||||
|
Loading
|
|||||
Say you have a design that looks like a four group ANCOVA, but your dependent variable is a 0/1 variable. In such a case, running a normal ANCOVA is not really appropriate since the variable is 0/1, so instead you use probit. You code the data using dummy codes (b1 through b3) to indicate the group effect and you have a covariate (cov1). You then run the probit as shown below
clear
input y grp b1 b2 b3 cov1, nolog
0 1 1 0 0 43
1 1 1 0 0 54
0 1 1 0 0 44
0 2 0 1 0 49
1 2 0 1 0 45
1 2 0 1 0 42
0 3 0 0 1 54
1 3 0 0 1 34
1 3 0 0 1 56
0 4 0 0 0 45
0 4 0 0 0 67
1 4 0 0 0 54
end
probit y b1 b2 b3 cov1, nolog
Probit regression Number of obs = 12
LR chi2(4) = 1.48
Prob > chi2 = 0.8300
Log likelihood = -7.5772029 Pseudo R2 = 0.0890
------------------------------------------------------------------------------
y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
b1 | -.1341121 1.115745 -0.12 0.904 -2.320932 2.052708
b2 | .689892 1.168133 0.59 0.555 -1.599607 2.979391
b3 | .7513844 1.106161 0.68 0.497 -1.416651 2.91942
cov1 | -.0188094 .0544449 -0.35 0.730 -.1255194 .0879006
_cons | .6009771 3.061287 0.20 0.844 -5.399034 6.600989
------------------------------------------------------------------------------
Then, if you want to get predicted probabilities for each cell, but adjusted for the covariate, you can use the adjust command below. Note that by(grp)is just giving you the probabilities for the four levels of grp.
adjust cov1, by(grp) pr ci
The output is shown below, with the predicted probabilities and the confidence intervals.
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: probit
Variables left as is: b1, b2, b3
Covariate set to mean: cov1 = 48.916667
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .325193 [.028601 .840207]
2 | .644598 [.125402 .970618]
3 | .667227 [.144304 .97293]
4 | .37482 [.028089 .898211]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
The results given above are for when cov1 is held constant at its mean value of 48.92. What if we wanted to see the adjusted probabilities when cov1 is held constant at 45 and at 50. This can be accomplished simply by setting the covariate to a given value using an equal sign as shown below.
adjust cov1=45, by(grp) pr ci
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: probit
Variables left as is: b1, b2, b3
Covariate set to value: cov1 = 45
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .352137 [.032091 .862538]
2 | .67164 [.149509 .973037]
3 | .69355 [.144947 .980785]
4 | .403056 [.020922 .938728]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
adjust cov1=50, by(grp) pr ci
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: probit
Variables left as is: b1, b2, b3
Covariate set to value: cov1 = 50
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .317891 [.026373 .83886]
2 | .636981 [.115327 .971249]
3 | .659791 [.139594 .97167]
4 | .36712 [.029351 .887135]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
By the way, this will work the same way if you are using logit instead of probit as shown below.
logit y b1 b2 b3 cov1, nolog
Logistic regression Number of obs = 12
LR chi2(4) = 1.47
Prob > chi2 = 0.8313
Log likelihood = -7.5808132 Pseudo R2 = 0.0886
------------------------------------------------------------------------------
y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
b1 | -.2337632 1.867017 -0.13 0.900 -3.893049 3.425523
b2 | 1.108887 1.911031 0.58 0.562 -2.636666 4.854439
b3 | 1.199861 1.809699 0.66 0.507 -2.347084 4.746805
cov1 | -.0290209 .0869278 -0.33 0.738 -.1993962 .1413545
_cons | .9010222 4.897605 0.18 0.854 -8.698107 10.50015
------------------------------------------------------------------------------
adjust cov1, by(grp) pr ci
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: logit
Variables left as is: b1, b2, b3
Covariate set to mean: cov1 = 48.916667
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .32031 [.039684 .843119]
2 | .643435 [.131805 .955455]
3 | .664024 [.14934 .956989]
4 | .373184 [.042341 .889099]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
adjust cov1=45, by(grp) pr ci
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: logit
Variables left as is: b1, b2, b3
Covariate set to value: cov1 = 45
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .345545 [.044664 .856378]
2 | .669067 [.154548 .957193]
3 | .688892 [.151591 .96484]
4 | .400131 [.034988 .924653]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
adjust cov1=50, by(grp) pr ci
-------------------------------------------------------------------------------------------------------------
Dependent variable: y Command: logit
Variables left as is: b1, b2, b3
Covariate set to value: cov1 = 50
-------------------------------------------------------------------------------------------------------------
----------------------------------------------
grp | pr lb ub
----------+-----------------------------------
1 | .313505 [.037237 .843556]
2 | .63619 [.122668 .956275]
3 | .656974 [.144724 .955903]
4 | .365859 [.043598 .879546]
----------------------------------------------
Key: pr = Probability
[lb , ub] = [95% Confidence Interval]
UCLA Researchers are invited to our Statistical Consulting Services
We recommend others to our list of Other Resources for Statistical Computing Help
These pages are Copyrighted (c) by UCLA Academic Technology Services