Sociology 405/805

January 8, 2004

 

Review and Introduction

 

1. Production of Data

 

Use existing data sources in this class. Be familiar with variables and how they are defined, and with the methods of data collection. For this class, we will have to work with these data as they come to us, but when working with these data, should be aware of definitions and methods used, and what procedures were used to obtain the data.

 

        Data are produced not just collected.

        Social research issues, theory, approach.

        Definition of variables theoretical and operational.

        Questions used and potential answers.

        Organization of responses when data are presented.

        Sampling procedures.

      Population or sample.

      Non-probability or probability.

      Random, stratified, cluster, multistage.

      Interview, questionnaire, administrative data.

        Errors in data non-sampling and sampling.

 

        Integrate data production with statistical procedures to be used, if possible.

        Use statistical analysis of previous data sets to improve data production for subsequent projects.

        Replicate existing studies or use questions from existing studies for comparative purposes.

 

2. Types of Measurement

 

a. Discrete or Continuous

 

        Discrete number of possible values can be counted.

        Continuous cannot count all possible values; possible values can be matched with some portion of a line segment.

 

b. Level of Measurement

 

        Nominal Can classify values into categories; name or number them. Sex, ethnicity.

        Ordinal Values can be ordered or ranked as less than, greater than, or equal to. Order of finish, Likert-scale attitudes.

        Interval Differences or intervals are meaningful; equal numerical differences represent equal magnitudes; well-defined unit of measure. Height, weight, time, age, income.

        Ratio Ratios of values meaningful; non-arbitrary 0 point. Height, weight, income, age.

 

        Levels of measurement are hierarchical. That is, all scales are nominal. Ordinal scales are also nominal. Interval scales are both nominal and ordinal. Ratio scales are nominal, ordinal, and interval.

        Most interval scales are also ratio. Temperature may be only interval and not ratio.

        These levels of measurement determine the type of summary statistic that can be calculated and the type of statistical analysis that can be used.

        Where possible, construct variable in order to measure it at the highest possible level.

        Attitudes, opinions and many psychological variables are measured only at the ordinal level, but statistical analysis appropriate for interval or ratio level scales is commonly used.

        Some statistical methods appropriate only for interval or ratio level scales can use nominal or ordinal level scales through appropriate reconstruction of variables e.g. dummy variables.

 

3. Descriptive Statistics

 

a. Distributions. Frequency, percentage, proportional distributions. Organization of data into categories. Need nominal scale only.

b. Positional Measures. Percentiles, deciles, quintiles, quartiles, median. Need ordinal scale.

c. Central Tendency. Mode, median, mean. Mode for nominal scale, median for ordinal, mean for interval and ratio.

d. Variation. Variation, variance, standard deviation. Assume interval or ratio level scales. Use variation for ANOVA. For ordinal scales, interquartile range or other measures based on positional measures can be used.

e. Measures of Association. Lambda, phi, V, Q, tau. Covariation and correlation coefficients.

f. Regression. Regression coefficients and regression equation.

g. Standardization. Z-values and beta coefficients.

 

        Each statistic provides a particular summary view of the distribution of the data.

        Use statistics appropriate to level of measurement and anticipated use of the data.

        Interpretation of statistics.

        Where data come from samples or where the data make inferences concerning the influence of particular variables, the variability of the estimate must be considered. Use of interval estimates and hypothesis tests.

 


4. Probability and Sampling Distributions

 

        Classical, frequency, subjective interpretations of probability.

        Independence and dependence of events.

        Random variables and expected values.

        Probability distributions Normal, t, F, and chi-square distributions.

        Sampling

Representative

Random

Sampling Distributions

Stratified, cluster, multistage samples

Sampling error

        Models

Probabilistic, not deterministic

Multivariate

Description and explanation

 

5. Inferential Statistics

 

a. Sampling Error

b. Interval Estimates

c. Hypothesis Tests

 

6. Other Issues

 

 


7. Statistical programs SPSS and Minitab

 

 

 

Last edited January 10, 2004