Search Type Title Author/Investigator Search Advanced Search

Table of Contents | Previous | Next

Appendix C: Comparison of Fall 2002 Respondents and Non-Respondents

Non-response analysis for the child assessment and parent/primary caregiver interview survey instruments was conducted to determine the variables most correlated with response propensity for each instrument, i.e., to determine the factors that differentiate respondents from nonrespondents. Three definitions of response for each child were applied: 1) Child Assessment completed, 2) Parent Interview completed, or 3) both completed. The seven variables examined were the child’s age, gender, race/ethnicity, region, metro status, program option applied for, and treatment vs. control group assignment. The source of these variables was the random assignment roster that was completed by Westat field staff for every applicant seeking to enroll in the sampled Head Start centers; the roster, therefore, contained data for both respondents and nonrespondents.

Weighted and unweighted response rates for each of the three definitions are given in Tables C.1a and C.1b, where the weight is the child base weight that accounts for all stages of sampling and permits the treatment and control samples to represent the entire newly-entering Head Start population. The tables show that the response rates for children assigned to the non-Head Start control group are below those for children assigned to the Head Start treatment group, by about 10-12 percentage points on average. Most of the difference is due to the greater difficulty in locating control group children, since they do not attend Head Start. It is also interesting to note that although the overall average response rate is approximately 80 percent, there is wide variation among grantee/delegate agencies. The response rates at the great majority of grantee/delegate agencies range between 60 and 100 percent, while a small number of grantee/delegate agencies (6-10) have rates ranging from 60 percent to 25 percent for the fall 2002 data collection.34

To analyze differences between respondents and nonrespondents, bivariate chi-square tests of association between response status and each of the variables age, gender, region, metro status, program option applied for, and race/ethnicity were performed separately for the treatment and control groups. A logistic regression model was also run with the response status as the dependent variable (1=respondent, 2=nonrespondent) and the seven variables mentioned above as predictors. Both chi-square and logistic regression analyses were done using the child base weights and WesVar software to correctly account for the weights and the stratification and clustering in the Head Start Impact Study sample design. To prevent extremely large child weights from having undue influence on the results, 29 child base weights exceeding 1000 (0.61% of the sample) were truncated at 1000 and the analyses were rerun using this trimmed weight. The p-values for the chi-square tests are given in Table C.2 below and for the logistic regression analysis in Tables C.3 and C.4.

The Chi-square tests showed that aside from treatment vs. control status, only program option applied for within the control group was consistently significant across the three different response definitions. Response rates for control children applying for the Full-Day option are lower than for those applying for the Part-Day, which may be an indication that applicants to the Full-Day option tend to be parents who work full-time and are less available for the Parent Interview or Child Assessment. It may also indicate a higher refusal rate from those who wanted a full time option and were assigned to the control group. It is also worth pointing out that program report on this particular variable was inconsistently worded across sampled centers. As these are preliminary results, further analysis is ongoing. Head Start region was significant with the original weight but not the trimmed weight, which implies that the significance might have been caused by large outlier weights, and was not a true difference in response rates among the regions. Response rates also differed by gender for Child Assessment completes with the original weight but not the trimmed weight, again suggesting that the difference was caused by a few influential weights having extremely high values. As can be seen in Tables C.1a and C.1b, response rates for girls are only slightly higher than for boys.

In the logistic regression analysis, as expected, the control/treatment indicator variable was highly significant in predicting the propensity to respond for all three response definitions. Program option and gender were moderately significant predictors with the original child base weight, but not for the trimmed weights.

Table C.1a. Fall 2002 Parent Response Rates Based On Selected Characteristics For Children In The Head Start Treatment Group*

Child characteristics	Response rates based on:
	Child Assessment completes		Parent Interview completes		Both instruments complete
	Unweighted	Weighted	Unweighted	Weighted	Unweighted	Weighted
TREATMENT GROUP	86.04	86.82	86.31	87.52	84.66	85.73
Child's Gender
Boys	84.89	85.20	85.71	86.66	83.71	84.28
Girls	87.27	88.64	87.07	88.60	85.75	87.39
Agency Region
Northeast	77.51	80.56	77.83	82.01	74.93	78.63
Midwest	88.25	86.86	88.02	86.48	86.87	85.48
South	89.35	88.61	89.63	89.63	88.39	87.81
West	88.63	89.77	89.27	89.96	87.98	89.16
Agency Metro Status
MSA	86.04	87.12	86.18	87.78	84.55	86.04
NonMSA	86.07	85.88	86.87	86.72	85.11	84.77
Program Option Applied For
Full-Day	84.08	84.86	83.61	85.36	82.10	83.29
Part-Day	88.37	88.76	88.82	89.27	87.34	87.88
Both full day and part-day	71.43	75.51	79.59	82.96	71.43	75.51
Other**	85.34	87.41	87.93	88.33	85.34	87.41
Child's Race-ethnicity
Hispanic	84.97	85.63	86.43	87.27	83.87	84.84
Black	86.23	87.35	85.43	87.35	84.36	85.98
White	87.96	88.90	87.70	88.92	86.60	87.53
Other	78.95	78.50	79.31	78.50	78.95	78.50
Child's Age
Age 3	86.89	87.05	87.09	88.08	85.30	85.69
Age 4	85.01	86.56	85.33	86.89	83.87	85.78
Parent Language
English	86.84	87.86	86.54	88.05	85.28	86.60
Spanish	85.09	85.93	86.76	87.80	84.06	85.17
Other**	63.89	55.46	55.46	64.86	63.89	55.46

* Data Source: Roster information used at time of random assignment ** These categories have very small cell sizes.

 

Table C.1b. Fall 2002 Parent Response Rates Based On Selected Characteristics For Children In The Non-Head Start Control Group *

Child characteristics	Response rates based on:
	Child Assessment completes		Parent Interview completes		Both instruments complete
	Unweighted	Weighted	Unweighted	Weighted	Unweighted	Weighted
CONTROL GROUP	73.74	76.27	75.62	77.43	73.45	75.71
Child's Gender
Boys	72.27	74.32	74.59	76.25	72.16	73.91
Girls	75.22	78.02	76.67	78.47	74.73	77.30
Agency Region
Northeast	68.10	72.41	69.76	72.48	67.39	70.03
Midwest	75.09	75.27	77.14	77.00	75.36	75.50
South	72.98	74.95	75.32	76.73	72.73	74.90
West	82.46	82.32	83.38	83.18	85.15	82.13
Agency Metro Status
MSA	72.82	75.69	74.93	76.92	72.45	74.94
NonMSA	77.71	78.08	78.57	79.01	77.71	78.08
Program Option Applied For
Full-Day	67.46	69.46	69.46	70.94	66.83	68.22
Part-Day	78.70	81.35	80.70	82.52	78.76	81.36
Both full day and part-day	67.74	65.07	64.52	61.98	64.52	61.98
Other**	80.49	82.77	81.71	83.19	80.49	82.77
Child's Race-ethnicity
Hispanic	74.48	77.09	76.72	78.83	74.07	76.72
Black	71.58	75.60	72.86	74.88	70.86	73.82
White	74.58	74.82	76.55	76.81	74.86	75.00
Other	78.57	86.41	80.95	86.96	78.57	86.41
Child 's Age
Age 3	73.58	75.88	76.08	77.26	73.13	74.89
Age 4	73.95	76.72	75.03	77.63	73.83	76.64
Parent Language
English	73.07	74.74	75.08	76.35	73.02	74.79
Spanish	75.78	79.73	77.56	80.56	75.00	78.43
Other**	64.71	72.98	58.82	55.43	58.82	55.45

* Data Source: Roster information used at time of random assignment ** These categories have very small cell sizes.

 

Table C.2: P-Values for Chi-Square Tests of Association between Response Status and Child Variables, Calculated Separately for Treatment and Control Groups*

Child characteristics by Control/treatment group	Response status based on:
	Child Assessment completes		Parent Interview completes		Both completes
	Original weight	Trimmed weight **	Original weight	Trimmed weight **	Original weight	Trimmed weight **
Child Gender
treatment	.034	.450	.231	.423	.121	.492
control	.176	.220	.469	.465	.337	.237
Agency Region
treatment	.016	.209	.058	.179	.030	.125
control	.120	.429	.113	.432	.155	.362
Agency Metro Status
treatment	.729	.733	.740	.694	.802	.740
control	.663	.667	.681	.676	.621	.556
Program Option Applied For
treatment	.515	.573	.588	.592	.609	.532
control	.003	.016	.007	.042	.014	.023
Child Race-ethnicity
treatment	.635	.677	.631	.764	.826	.768
control	.526	.686	.403	.620	.612	.675
Child Age
treatment	.859	.837	.516	.700	.954	.972
control	.729	.793	.822	.921	.560	.632

* Data source: Roster information used at time of random assignment. ** Note: bold print indicates statistically significant difference.

 

Table C.3: P-Values for T-Tests of Significance for Logistic Regression Parameters for Combined Treatment and Control Groups*

Predictor Variables	Response status based on:
	Child Assessment completes		Parent Interview completes		Both completes
	Original weight	Trimmed weight **	Original weight	Trimmed weight **	Original weight	Trimmed weight **
Intercept	0.003	0.003	0.004	0.018	0.010	0.002
Child's Age is 3	0.247	0.248	0.306	0.487	0.501	0.404
Child's Gender is female	0.052	0.314	0.062	0.141	0.168	0.310
Agency Region 1 -Northeast vs. West	0.048	0.069	0.267	0.390	0.056	0.048
Agency Region 2 -Midwest vs. West	0.182	0.185	0.641	0.710	0.348	0.256
Agency Region 3 -South vs. West	0.285	0.289	0.965	0.967	0.552	0.465
Agency Metro Status -Rural	0.812	0.871	0.442	0.527	0.869	0.891
Program Option Applied For 1 -Full-day vs. Other	0.347	0.380	0.855	0.895	0.402	0.337
Program Option Applied For 2 -Part-day vs. Other	0.882	0.890	0.446	0.528	0.954	0.952
Program Option Applied For 3 -Full- or Part-day vs. Other	0.421	0.415	0.817	0.909	0.509	0.413
Child Race-Ethnicity - Hispanic vs. White	0.721	0.694	0.957	0.942	0.768	0.692
Child Race-Ethnicity - Black vs. White	0.331	0.540	0.994	0.980	0.629	0.659
Child Race-Ethnicity - Other vs. White	0.925	0.974	0.132	0.268	0.903	0.933
Child is in treatment group	0.000	0.006	0.000	0.000	0.000	0.003

* Data Source: Roster information used at time of random assignment. ** Note: bold print indicates statistically significant difference.

 

Table C.4. P-Values for F-Tests of Overall Significance for Logistic Regression Variables for Combined Treatment and Control Groups*

Predictor Variables	Response status based on:
	Child Assessment completes		Parent Interview completes		Both completes
	Original weight	Trimmed weight **	Original weight	Trimmed weight **	Original weight	Trimmed weight **
Overall Fit	0.000	0.000	0.000	0.000	0.000	0.000
Child's Age (2)	0.247	0.248	0.132	0.268	0.501	0.404
Child's Gender (2)	0.052	0.314	0.306	0.487	0.168	0.310
Agency Region (4)	0.256	0.316	0.226	0.451	0.261	0.250
Agency Metro Status (2)	0.812	0.871	0.965	0.967	0.869	0.891
Program Option Applied For (4)	0.029	0.114	0.046	0.173	0.093	0.114
Child's Race-Ethnicity (4)	0.701	0.797	0.994	0.998	0.936	0.910
Treatment/Control (2)	0.000	0.006	0.000	0.000	0.000	0.003

* Data Source: Roster information used at time of random assignment. ** Note: bold print indicates statistically significant difference.

---

34The actual number of grantee/delegate agencies with these low rates differs slightly depending upon the respondent and type of instrument administered. (back)

 

Table of Contents | Previous | Next