Hi, so for my research project I am investigating effect a blood pressure diagnosis has on alcohol consumption frequency and physical activity. I have constructed 2 different models for both and I will be using xtologit command as I have panel data from 3 waves.
Please see my output below. My DV is alcohol_freq which is an ordinal variable of 4, which measures days per week the individual drinks. (0-1 1-2- 2 3-5 -3 6-7 -4).
My independent variables are gender (1 is female), net financial wealth, education level, waves 4 and 6 (baseline wave is 2) and alcohol frequency at baseline.
So from my output I guess I can conclude as time went on (waves4 and 6) the no. of days of drinking per week decreased for my sample, and this became stat significant in wave 6.
However I am confused by the positive coefficients for baseline alcohol, wealth group, education. Does this mean having a higher net wealth would increase frequency of drinking per week over the study? Or I am interpreting this wrong?
Any help would be appreciated. Thanks
Please see my output below. My DV is alcohol_freq which is an ordinal variable of 4, which measures days per week the individual drinks. (0-1 1-2- 2 3-5 -3 6-7 -4).
My independent variables are gender (1 is female), net financial wealth, education level, waves 4 and 6 (baseline wave is 2) and alcohol frequency at baseline.
Code:
xtologit alcohol_freq ragender i.wealth_group i.raeduc_e i.baselinealch _Iwave_4 _Iwave_6 if bp==1, vce(robust)
Random-effects ordered logistic regression Number of obs = 1,516
Group variable: idauniq Number of groups = 410
Random effects u_i ~ Gaussian Obs per group:
min = 1
avg = 3.7
max = 4
Integration method: mvaghermite Integration pts. = 12
Wald chi2(12) = 389.23
Log pseudolikelihood = -1188.8292 Prob > chi2 = 0.0000
(Std. Err. adjusted for 410 clusters in idauniq)
------------------------------------------------------------------------------------------
| Robust
alcohol_freq | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
ragender | -.5086046 .175747 -2.89 0.004 -.8530623 -.1641468
|
wealth_group |
2 | .1466781 .2819258 0.52 0.603 -.4058864 .6992425
3 | .1736882 .3064263 0.57 0.571 -.4268963 .7742727
4 | 1.099372 .3818805 2.88 0.004 .3508996 1.847844
|
raeduc_e |
3.high-school graduate | .2751227 .2454674 1.12 0.262 -.2059845 .75623
4.some college | .4316852 .2408849 1.79 0.073 -.0404405 .9038109
5.college and above | .1977728 .2608726 0.76 0.448 -.3135281 .7090737
|
baselinealch |
2 | 3.437855 .2885609 11.91 0.000 2.872286 4.003424
3 | 5.746778 .3989152 14.41 0.000 4.964919 6.528638
4 | 8.99715 .5110363 17.61 0.000 7.995537 9.998763
|
_Iwave_4 | -.078177 .1376317 -0.57 0.570 -.3479302 .1915762
_Iwave_6 | -.6799089 .1630516 -4.17 0.000 -.9994842 -.3603337
-------------------------+----------------------------------------------------------------
/cut1 | 1.375678 .4032719 3.41 0.001 .5852797 2.166077
/cut2 | 4.567226 .4576469 9.98 0.000 3.670255 5.464198
/cut3 | 7.281305 .5016476 14.51 0.000 6.298093 8.264516
-------------------------+----------------------------------------------------------------
/sigma2_u | 1.255873 .2308611 .8759377 1.800603
------------------------------------------------------------------------------------------
Code:
So from my output I guess I can conclude as time went on (waves4 and 6) the no. of days of drinking per week decreased for my sample, and this became stat significant in wave 6.
However I am confused by the positive coefficients for baseline alcohol, wealth group, education. Does this mean having a higher net wealth would increase frequency of drinking per week over the study? Or I am interpreting this wrong?
Any help would be appreciated. Thanks