biostats2

Results

As specified in the methods sections a separate linear regression model was run for each output variable. Originally, all five explanatory variables were used in these models. For each model, significant variables were indeed found when adjusting for all five variables. However, the Household status variable proved to be particularly insignificant because its p-value was >.9 in both models. After consulting with Dr. Wilson, we eliminated this fifth variable. He had advised to use only variables that seemed to be making an impact on our output variables.

Thereafter, the model was run adjusting for only four variables. We found that the perceived need for public school quality was driven by two variables. Both Race/Ethnicity and Delegate Region had significant p-values; .0189 and <.0001 respectively. The perceived need for medical care was found to be driven by only Delegate Region. Its p-value was .0002.

As seen by these results, Employment Status, Income nor Household status made an impact on the output variables. The implications of these results are only partly explainable. One of the requirements for participation in the Head Start program is a income near the poverty line. As a result, all respondents may have similar incomes and therefore this variable did not impact the preception of needs. The monthly income had a mean of $1453 and a median of $1000. Though the standard deviation was $3592 it is clear that this is because some respondants reported an annual salary instead of a monthly one. The historgram shows a better picture of the income distribution. A further study to look at the distribution of income might reveal why income did not drive the output variables.

A similar focussed study on Employment status might explain why it did not make an impact. It should be noted that >82% of the respondants were considered employed. Lastly, >67% of households were Two Parent households. This may imply that because the distribution was not normal, it did not drive the output variables. Analysis has revealed some interesting biases within the program that should be taken into account for any further studies.

• linear regression model was run twice
• household status was found to be insignificant
• give p-values
• recommended to focus only on significant variables
• second time model was run with 4 variables
• public school quality

two variables Race = .0189-- Region < .0001

• medical care

one variable region = .0002

• as seen the other variables do not have an impact
• conclusion, it is clear that these do have an impact on perceived need