Prev: An unsuccessful attempt to parse the SAS logfile for a WARNING-message
Next: how to filter observations
From: RAJ on 22 Mar 2010 21:26
Hi there,

This is the first time am trying to use sas and need some help doing
my Multi level analysis. I usually use SPSS and other statistics
software to do my analysis. It will be great if i can get some help
doing some coding. The problem is defined below. My boss is being a
pain and wants me to use SAS.

The problem is defines below

Multi-level Analysis of Language Proficiency

The benefit of using a multilevel analysis is so that student and
school level variance can be identified and reported. In analyses of
L1 proficiency, the unconditional model will indicate what percentage
of variation in L1 proficiency is due to student level variation and
school level variation. Similarly in analyses of L2 proficiency, the
unconditional model will indicate what percentage of variation in L2
proficiency is due to student level variation and school level
variation. To the best of my knowledge, that is the first step in a
multilevel analysis and the first result to be reported. (Note: I
call this the baseline model or intercept only model).
Student level variables and school level variables were sent as
below.
_______________________________________________________________________
Outcome Variables
 L1study_sc—L1 proficiency
 L2study_sc—L2 proficiency

Student-Level Variables
 Gender—Male, Female
 Year—dummy variable to be created (for some students L1year is the
same as L2year and for others L1year is different from L2year)
 L1study_sc—please note that L1study_sc is a student level predictor
when estimating L2 proficiency
 L2study_sc—please note that L2 study_sc is a student level predictor
when estimating L1 proficiency
 L2Stud_type—EN,ESL

School-Level Variables
 L1Schl_typeID—L1 school types/sectors
 L1Schl_DID—participating L1 schools
 L1Prg_model—REG, SA
 L2Schl_typeID—L2 school types/sectors
 L2Schl_DID—participating L2 schools
_____________________________________________________________________

I would have a two level, i.e. student level and school level
analysis. In other words, I would have students nested under
schools. Specifically,
• In L1 analyses, I would have students nested under L1 schools. I
would introduce L1Schl_type or L1Prg_model as school level variables
to see whether controlling for L1school level variation, L1sector or
L1program model makes a difference.

• In L2 analyses, I would have students nested under L2 schools. I
would introduce L2Schl_typeID as school level variable to see whether
controlling for L2 school variation, L2schl_type makes a difference.



L2 proficiency analysis should examine whether there is empirical
evidence to support the theoretical postulate that students’ L1
proficiency (L1study_sc) is a statistically significant predictor of
L2 proficiency (L2study_sc) controlling for variation across L2
schools (L2Schl_DID), and if so, examine whether the relationship
between L1 proficiency (L1study_sc) and L2 proficiency (L2study_sc)
differs
• across L2 school_types/sectors (L2Schl_typeID);
• for EN and ESL students(L2Stud_type);
• for male and female students (Gender);
• for students who sat for the L1 and L2 matriculation exams during
the same year and in different years (Year).

WRITTEN AS AN EQUATION

L2Study_sc= L1Study_Sc + Gender + Year + L2Stud_type+ L2Schl_type/
sector

(Please note that L1Study_Sc, Gender, Year, L2 Stud_type should be
introduced at Student level, and L2 Schl_type/sector variable should
be introduced at School level.)

Note: The interpretation of L2 analyses should start by indicating
the percent of variance explained by student level and school level
variation (i.e. the unconditional model).
The table summarizing findings of L2 analyses should contain the
effect of each of the five variables, i.e. L1Study_Sc, Gender, Year,
L2Stud_type, L2Schl_type/sector.

L1 proficiency analysis should examine whether there is empirical
evidence to support the theoretical postulate that students’ L2
(L2study_sc) is a statistically significant predictor of L1
proficiency (L1study_sc) controlling for variation across L1 schools
(L1Schl_DID)?

• across L1program models (L1_Prgmodel);
• across L1 school_types/sectors (L1Schl_typeID);
• for EN and ESL student types (L2Stud_type);
• for male and female students (Gender);
• for students who sat for the L1 and L2 exams during the same year
and in different years (Year).


WRITTEN AS AN EQUATION

L1Study_sc= L2Study_Sc + Gender + Year + L1Schl_type/sector
L1Study_sc= L2Study_Sc + Gender + Year + L1_Prgmodel

Please note:
 L1Study_Sc, Gender, Year, L2 Stud_type should be introduced at
Student level, and L1Schl_type/sector or L1_Progmodel variable should
be introduced at School level
 Due to overlap, L1Schl_type/sector and L1_Prgmodel can’t be used in
the same analysis that is why I am sending you two equations
 And if so, examine whether the relationship between L1 proficiency
(L1study_sc) and L2 proficiency (L2study_sc) differs

The interpretation of L1 analyses should start by indicating the
percent of variance explained by student level and school level
variation (i.e. the unconditional model).
The table summarizing findings of L1 analyses should contain the
effect of each of the five variables, namely L2Study_Sc, Gender, Year,
L1Schl_type/sector, L1_Prgmodel

Note: In L1 analyses we should have about 20 schools and in L2
analyses we should have about 100 schools.
From: Patrick on 23 Mar 2010 08:05
Sounds as if your boss is a teacher.....
From: RAJ on 23 Mar 2010 15:20
On Mar 23, 7:05 am, Patrick <patrick.mat...(a)gmx.ch> wrote:
> Sounds as if your boss is a teacher.....

WEll i guess nobody wants to help :(
From: Patrick on 24 Mar 2010 09:42
RAJ

May be you demonstrate first that you've put some effort and thinking
in solving the problem and when you get stuck in some step you come
back and ask an informed and specific question.

Don't expect people to do your assignements for you.

Cheers
Patrick
 | 
Pages: 1
Prev: An unsuccessful attempt to parse the SAS logfile for a WARNING-message
Next: how to filter observations