• If any of the questions are not completely answered
  
• If the pattern of the responses indicates that the respondent did not understand the survey
• The responses show very little variance
• If it is answered by someone who does not qualify for participation

• Return to the field and gather the complete response from that individual
• Assign missing values to unsatisfactory responses
• Discard the entire questionaire with unsatisfactory responses

• Substitute a neutral value (i.e. the mean response for that variable from all of the other respondents)
• Substitute an imputed response (i.e. the respondents' pattern of response to other questions are used to calculate a suitable response to the missing question)
• Casewise deletion - cases, or respondents with any missing responses are deleted completely
  
• Pairwise deletion - instead of discarding all cases with any missing values, the researcher uses those cases with complete responses for each question when doing calculations

• t-Test
• two sample means t-Test
• Chi-Squared test
• ANOVA

• Regression analysis
• Conjoint analysis

• The researcher develops a theory about the data that was collected from N individuals
• The one-way chi-squared analysis tests how close the data is to the theory
• The theory can come from historical data

Because the Chi-Square statistic is a function of the deviation of the observed counts from the theory, a larger Chi-Square statistic means that there is a big difference between the theory of the data and the actual data.

When the chi-square statistic is large (i.e. bigger than the critical value) then reject the null hypothesis.  The actual data is new and the theory must be changed.

• Tests the mean of a single variable against a number
• Use with interval and ratio questions

• Paired means t-Test tests the mean for the same group to two different variables or activities
• Example: men's attitudes towards the Internet and men's attitudes towards technology in general
• Use with interval and ratio questions
  

• Use when comparing two different groups to the same variable or activity
• Example: men's opinions of the rec center and women's opinions of the rec center

• Use when testing 3 or more groups (as opposed to t-Testing which tests 1 to 2 groups) for the same variable or usage
• Example: Internet usage by income level, age groups, household size, etc.
  • Is mean internet usage the same across different income classes?
  • Null hypothesis would be: The mean internet usage in each class is equal to that of every other class.

• This tests for statistical independence
• Testing to see if one variable has nothing to do with another variable
• Null hypothesis: Variable 1 and Variable 2 are statistically independent
  • Rejecting the null hypothesis shows that the variables actually can affect one another.

It will show if there is a relationship between an interval and nominal variable by making a linear relationship between the two variables.

Should be completed prior to running a regression on variables.  Check for collinearity and make sure that the variables should even be included in the regression.

• Correlation numbers range from -1 to +1
• At or near 0 there is NO relationship between the two variables
• Low correlation is around +/- 0.50
• High correlation close to +/- 1.00

• When the independent variables (X's) are highly collinear together with each other
  
• When this happens do not include all of the X's in the regression model
  
• Choose which X to include

• The utility for a specific level of a particular attribute. 
• It designates how much that part of the product or service is worth to the consumer.
• Note that regression analysis usually is used for this step, the coefficients for each of the attributes that are given in the regression output ARE the part worths
  

• Note that importance is relative to the other attributes, show which attributes are most/least important
• Importance weights illustrate how much one attribute can contribute to the overall utility - i.e. an attribute cannot effect utility any more than it's maximum value of the importance weight

• Extremely useful tool in predicting the performance of new products and to help re-designing old products
• In addition, it can be used to determine price elasticity of demand and can help segment the market better

A₁(maximum part worth - minimum part worth)

Yes, IF:

• One of the attributes tested was price of the overall product
• Determine the range of part worths (max part worth - min part worth)
• That util range is the number of utils that the range in overall price is worth
• Divide the range of overall price by the range of part worths to get $X/1 util

• Testing too many attributes
• Attribute descriptions are vague or ambiguous
• No pre-test of attributes
• Completed too early, needs to be done way at the end of marketing research
  
• Convenience sample
• Incorrect attribute levels for competitors