Hi!
I am running a regression analyses for students across two grades that are using two learning products. (See attached .do file with copies of outputs - I cannot shared the data set here)
Usage data were not normal ( as is often the case in the real world) and could not be viably transformed (used ladder command in Stata) but since there was a sufficient sample size to be robust to the non-normal usage data, I used standardized beta weights to report the results.
For the overall group of students I ran a regression analyses, controlling for product1_usage product2_usage grade and pretest scores. I found a significant relationship between product1_usage and achievement as measured by the posttest, product1_usage predicted posttest results, R^2 = 0.67, F(4, 1238) =673.74, p = 0.000. After accounting for product2_usage, prior achievement and grade level, students who completed more product1_usage also had higher posttest performance.
Amount of product1_usage has a 0.08 standardized beta weight, this means that a 1 s.d. in product1_usage (10) is associated with a 0.08 s.d. increase in the posttest scale score. Completing 10 more "uses" of product1 is associated with an average increase of 7.4 points on the posttest.
I tried to run a similar analyses for subgroups of students who had FRL status (dichotomous) and subgroups of students by pretest quartiles. However, I don't think I can viably draw the same conclusions as I did with the overall group above. Also I need a method whose output is easily consumable by a lay person.
So, I was thinking about running an ANCOVA but I running into some trouble here: is there a way to get the matrix that compares the groups to each other in Stata? I have included my first stab at ANCOVA in the .do file also.
Thanks in advance for helping!
I am running a regression analyses for students across two grades that are using two learning products. (See attached .do file with copies of outputs - I cannot shared the data set here)
Usage data were not normal ( as is often the case in the real world) and could not be viably transformed (used ladder command in Stata) but since there was a sufficient sample size to be robust to the non-normal usage data, I used standardized beta weights to report the results.
For the overall group of students I ran a regression analyses, controlling for product1_usage product2_usage grade and pretest scores. I found a significant relationship between product1_usage and achievement as measured by the posttest, product1_usage predicted posttest results, R^2 = 0.67, F(4, 1238) =673.74, p = 0.000. After accounting for product2_usage, prior achievement and grade level, students who completed more product1_usage also had higher posttest performance.
Amount of product1_usage has a 0.08 standardized beta weight, this means that a 1 s.d. in product1_usage (10) is associated with a 0.08 s.d. increase in the posttest scale score. Completing 10 more "uses" of product1 is associated with an average increase of 7.4 points on the posttest.
I tried to run a similar analyses for subgroups of students who had FRL status (dichotomous) and subgroups of students by pretest quartiles. However, I don't think I can viably draw the same conclusions as I did with the overall group above. Also I need a method whose output is easily consumable by a lay person.
So, I was thinking about running an ANCOVA but I running into some trouble here: is there a way to get the matrix that compares the groups to each other in Stata? I have included my first stab at ANCOVA in the .do file also.
Thanks in advance for helping!
Comment