290 likes | 538 Views
STAT 8000: Supervised Statistical Consulting. Early Intervention Program in Morgan County, GA. Client: Dr. Wayne Myers Instructor: Dr. Jaxk Reeves By: Xiaofei Li, Allison Moore, and Minrong Song.
E N D
STAT 8000: Supervised Statistical Consulting Early Intervention Program in Morgan County, GA Client: Dr. Wayne Myers Instructor: Dr. Jaxk Reeves By: XiaofeiLi, Allison Moore, and MinrongSong
Client is Dr. Wayne Myers, Board of Education School Readiness Coordinator for Morgan County • Optional Pre-K at age 4 • Data for children entering Kindergarten in 2012 • Currently, Dynamic Indicators of Basic Early Literacy Skills (DIBELS) used as a standard for placing students in the Early Intervention Program (EIP) introduction
Can the measures taken in Pre-K be used to predict the Kindergarten DIBELS scores? 2. Do Pre-K measures perform better than the DIBELS scores when it comes to predicting EIP enrollment? RESEARCH QUESTIONS
DIBELS are a set of one minute fluency measures to monitor literacy skills • Phonemic awareness, vocabulary, comprehension, alphabetic principles • Square root of DIBELS • EIP is a program that • provides additional • resources to at risk • students • IE: Indicator of EIP Response VARIABLES
DEMOGRAPHIC VARIABLES ( ) Created or client calculated variable
Get Ready to Read (GRTR) is a test designed to measure reading ability • Number of flash cards a student can correctly identify PRE-SCHOOL MEASURES
Literacy rankings are assigned by the teacher (1-4) PRE-SCHOOL MEASURES
16 lines contained missing observations • NR, NT, 0, or blank cells were changed to “.” Missing data
GRTR_SCORE > 20 is considered high • DIBELS = 0 is really a zero Data summary
Values greater than .800 cause some concern • Highly correlated pairs should not be included simultaneously in a model CORRELATION MATRIX
STEPS: • Regress individual variables separately on DIBELS • Calculate their t-values to find how useful individual variables are in predicting DIBELS • Build models using this knowledge • Compare models using AIC and BIC 1. Can the measures taken in pre-k be used to predict dibels?
Not significant: • GENDER (M) • AGE • FERST (Y) T-VALUES FOR PREDICTING DIBELS • *Ranked by absolute t-value. All p-values < .0001
Interpret a model with interaction between two continuous variables need standardization, because: • Make sure all the variables contribute evenly to a scale • To facilitate interpretation STANDARDIZATION
RESULTS – BEST MODELS • *** p-value < .01 • ** p-value < .05 • * p-value < .10 • With average GRTR_SCORE, average LMEAN, & IS=0: • Model 2 suggests DIBELS = 17.95 • Model 3 suggests DIBELS = 18.38
STEPS: • Use simple logistic regression on individual variables to determine importance in predicting EIP enrollment • Calculate their Chi-square values to find how useful individual variables are in predicting EIP • Build models using this knowledge • Compare models using AIC and BIC 2. pre-k measures v.s. Dibels at predicting eip enrollment
Not significant: • GENDER (M) • FERST (Y) • Negative coefficients Lower probability of EIP enrollment X2-VALUES FOR PREDICTING EIP • *Ranked by chi-square values. All p-values < .0001
Results – bEST MODELS • Under the null model: • *** p-value < .01 • ** p-value < .05 • * p-value < .10
81.8% • DIBELS = 10 Results – MODEL 6 • 0.16%
Black and Hispanic without SUPPORT Asian, White, and Multi-Racial with SUPPORT Results – Probability prediction chart • *Holding the GRTR_SCORE constant
DIBELS to Predict EIP Overall 84.71% correct • Best Model: Pre-K Measures to Predict EIP Overall 86.47% correct Results – PREDICTING EIP
CONCLUSION • Predicting DIBELS: • GRTR_SCORE, LMEAN, and their interaction are found to be significant in predicting DIBELS • DIBELS scores increase as GRTR and Literacy scores increase • Predicting EIP Enrollment: • IS, IOH, GRTR_SCORE, LMEAN and GRTR_SCORE*LMEAN are significant in predicting the probability of enrolling in EIP • This model predicts accurately 86.47% of EIP results • Using Pre-K measures is only marginally better than the model that uses only DIBELS scores to predict EIP
LESSONS LEARNED • Ask client about outside or unknown factors: • 7 students with DIBELS=0 did not enroll in EIP • Students receiving support, often did not enroll in EIP • Try something new: • Standardizing variables because of interaction term • Check the data: • Client’s definition of ‘missing data’ or ‘0’ • ‘Y’ versus ‘y’
”Early Intervention Program.” Georgia Department of Education, n.d. Web. 17 Apr. 2013. <www.doe.k12.ga.us/Curriculum-Instruction-and-Assessment/Curriculum- and-Instruction/Pages/Early-Intervention-Program.aspx>. Good, Roland H., III, and Ruth A. Kaminski. ”Dynamic Indicators of Basic Early Literacy Skills.” Dynamic Measurement Group, 2009. Web. 17 Apr. 2013. <dibels.org/dibels.html>. ”Pre-K.” Bright From the Start. Georgia Department of Early Care and Learn- ing, 2013. Web. 10 Apr. 2013. <decal.ga.gov/Prek/PreKHome.aspx>. ”Program Overview.” Ferst Foundation for Childhood Literacy, n.d. Web. 17 Apr. 2013. <www.ferstfoundation.org/>. ”49% Near Half of Morgan County’s Students Are Part of the Free and Reduced- Price Meals Program” Morgan County Citizen, 15 Dec. 2011. Web. 17 Apr. 2013. <www.morgancountycitizen.com/?q=node/19032>. references