Testpaper4

Sargis Hovhannisyan

Multiple Regression Lab

Study Design

1. Comment on the study design. How were the children in the study selected? Can you generalize the results of the study to any well-defined population? Can the study be used to demonstrate a causal relationship between breast milk consumption in the early weeks of life and cognitive ability of the children at 7-8 years measured by IQ tests? Explain briefly.

2. How important for the final conclusions in the study was the fact that the infants were being fed mother’s milk by tube rather than with the process of breastfeeding? Explain briefly.

3. The variables SOCIAL and EDUC specify which of the 4 or 5 categories the mother belongs to. What is the benefit of treating these variables as numerical variables rather than categorical variables? What assumption are we making about the relationship between the categories for these two variables (from 1 to 4 for SOCIAL and from 1 to 5 for EDUC)?

Exploratory Data Analysis

4. Obtain a matrix of scatterplot of the first five variables (IQ and the four predictors, WEEKS, SOCIAL, EDUC and RANK. Paste the scatterplot into your report.

(a) Comment on the relationship between IQ and each of these four predictors. In particular, look for any non-linearity in the relationships and potential outliers. Does it look like a linear model is appropriate for describing the relationship between IQ and any of these four predictors on the original scales? Note: If you have any doubt then produce individual scatterplots for IQ vs. each of these four predictors. Don’t paste the individual scatterplots into your report.

IQ vs weeks plot shows a negative correlation and is the strongest one of all. IQ vs social plot shows a negative weak association. There is a positive weak association between IQ and educ, and it looks like there is no noticeable association between IQ and rank. It looks like a linear model is appropriate for describing the relationship between IQ and any of...