|
|
Assessing dropout in a longitudinal s... |
|
Message/Author |
|
J Dennis posted on Monday, March 03, 2014 - 10:43 am
|
|
|
Hello, I am trying to obtain descriptive statistics related to dropout in a longitudinal study as per e.g. 11.2 in UG version 6. I can't seem to make it work, I was hoping you might be able to help me spot the problem in the code below. I was able to successfully run the code in 11.4 (pattern-mixture model) for my data but not 11.2. thanks in advance! Title: generating descriptive statistics for dropout in longitudinal study data: file S1v5.dat; variable: names = ID_2013A ALCDSMW1 ALCDSMW2 ALCDSMW3 ALCDSMW4 PGSIRW1 PGSIRW2 PGSIRW3 PGSIRW4 nPGSIRW1 nPGSIRW2 nPGSIRW3 nPGSIRW4 PGSIW1 PGSIW2 PGSIW3 PGSIW4 PGSICW1 PGSICW2 PGSICW3 PGSICW4 PGSID5W1 PGSID5W2 PGSID5W3 PGSID5W4 PGSID8W1 PGSID8W2 PGSID8W3 PGSID8W4 CIDIGAW1 CIDIGAW2 CIDIGAW3 CIDIGAW4 CIDIMDW1 CIDIMDW2 CIDIMDW3 CIDIMDW4 DRGR12W1 DRGR12W2 DRGR12W3 DRGR12W4 LEQW1 LEQW2 LEQW3 LEQW4 PSSTOTW1 PSSTOTW2 PSSTOTW3 PSSTOTW4 AGEW1 AGEW2 AGEW3 AGEW4 GENDER PG2W1 nPG2 BIGWIN BIGLOSS BIGWorL Region; usevar = nPGSIRW1-nPGSIRW4 ALCDSMW1-ALCDSMW4 d1-d3; missing = all(-99); DATA MISSING: NAMES = nPGSIRW1-nPGSIRW4; TYPE = DDROPOUT; BINARY = d1-d3; DESCRIPTIVE = nPGSIRW1-nPGSIRW4 | * ALCDSMW1-ALCDSMW4; ANALYSIS: TYPE = BASIC; PLOT: TYPE = PLOT2; SERIES = nPGSIRW1-nPGSIRW4(*); |
|
|
Please send input, output, data, and license number to Support. |
|
J Dennis posted on Monday, March 03, 2014 - 12:57 pm
|
|
|
OK will do --thank you. In a related question, how do I interpret the coefficients for the binary variables in the pattern-mixture model in e.g. 11.4? I have 4 wave data with a continuous outcome (time coding 0,1,2,3), so I created d1-d3 binary dropout indicators. Is it correct that, for example, the coefficients (the latent slope and intercept regressed on d2) for d2 represent the effect/bias (on the latent intercept and slope estimates) due to the difference between those who dropped out after w2 (did not complete w3) compared to those who completed w3? Thus a significant coefficient for d2 indicates a significant bias in growth factor estimates due to differences between dropouts and w3 completers? |
|
|
Hello, Is it possible to create binary dropout dummy indicators for three different sets of variables in the same "data missing" command? I've tried three plausible variations of the program below and received warnings each time. Thank you, Nick !V1 data missing: Names = fnc_1-fnc_8; TYPE=Ddropout; binary=fd1-fd7; Names = imr1-imr9; TYPE=Ddropout; binary= id1-id7; Names = sc1-sc9; TYPE=Ddropout; binary=sd1-sd7; !V2 data missing: Names = fnc_1-fnc_8 ; TYPE=Ddropout; binary=fd1-fd7 ; data missing: Names = imr1-imr9; TYPE=Ddropout; binary= id1-id7; data missing: Names = sc1-sc9; TYPE=Ddropout; binary=sd1-sd7; !V3 data missing: Names = fnc_1-fnc_8 | imr1-imr9 | sc1-sc9; TYPE=Ddropout; binary=fd1-fd7 | id1-id7 | sd1-sd7; |
|
|
There isn't a way. We only do it for one set of variables. The vertical bar (|) is only used in the DESCRIPTIVE option for the descriptive statistics. |
|
Back to top |
|
|