- StudyBlue
- California
- California State University - Long Beach
- Sociology
- Sociology 260
- Mechlinski
- MIDTERM #2

Brandi W.

Ordinary Least Squares Regression

(line of best fit - how far it goes up for how far it goes over)

takes the distance from the data point to the line & squares it.

plots line with smallest #

takes the distance from the data point to the line & squares it.

plots line with smallest #

Regression Coefficient

A one unit change in X is assoc with a ____ unit change in Y

Advertisement

What is most common technique in all of Soc Sci?

Regression Analysis

Pearson's r

measure. of assoc used with 2 scale variables

measured -1 to 1 strength being absolute distance from 0 not taking into sign, sign indicates strength

if using 2 variables regression. will give same #

measured -1 to 1 strength being absolute distance from 0 not taking into sign, sign indicates strength

if using 2 variables regression. will give same #

Cross-Tabulation

This is a fast and easy way to analyze and show data.

It does not include any tests of significance.

They can be as simple as 2X2, or much more complex, including control variables.

Another way of viewing bivariate data

Usually work for nominal or ordinal-type data

Allow for quantification of relationships

Cross-Tabulation

Definition

Definition

A matrix that shows the distribution on one variable for each category of a second variable

Bivariate Analysis

- The Analysis of 2 variables at a time
- major question - Why
- Primary Goal - Explanation
- Look for correlation, relationships & assoc (all same thing)

independent variable

- cause variable
- X axis
- Column

Dependent Variable

- Effect Variable
- Y axis
- Row

How to cross-tab 2 variables, make proper % and proper comparisons

input IV in column & DV in row be sure to check column percentages in cell display. Compare across & calculate down. If 10% diff worth talking about.

Lambda

- A measure of Assoc appropriate to nominal variables.

- Values (Strength) 0.00 to +1.00

1.00 = perfect assos

- Direction not applicable

Advertisement

Gamma

- A measure of assoc for ordinal variables.

- Strength -1.00 to +1.00

+1/-1 = perfect assoc.

- Direction: if both increase positive. if one increases & one decreased negative assoc.

Phi

Use when both variables are dichotomous (have only 2 categories) can use for nominal or ordinal

Descriptive Statistics

Describes the characteristics of a sample

Inferential Statistics

- making claims about the population based on testing the sample (Teacher def)

- making generalizations from samples to population (book def)

Normal Distribution

frequency is the highest near the mean value and decreases toward each extreme end of the range

Bell-Shaped curve

Bell-Shaped curve

Three types of Descriptive Statistics

- Measures of Central Tendency

- Measures of Dispersion

- Measures of Association

Statistical Significance

- Likelihood that an observed relationship between variables in a sample represents something other than sampling error. Usually by p<.05

Type I error

- 5% chance or less mistakenly rejecting the Null Hypothesis

- Mistakenly saying there is a relationship between variables when there isn't one

Type II Error

- Mistakenly accepting null Hypothesis.

- Saying there is a no relationship when really is one.

Level of significance

the standard for rejecting the null; typically set at 0.05

Level of Confidence

We are 95% sure that our outcome has not occurred by chance alone

Level of Risk

We are allowing ourselves 5% risk in making a Type 1 error

Positive Association / Direct Correlation

As one variable increases the other variable increases

Negative Association / Indirect Correlation

As one variable increased the other variable decreases

Interpreting Strength for Lambda, Gamma, Phi

The closer to 0 the weaker the assoc. The closer to 1.00 (either +/-) the stronger the assoc

Conditions of Causality

- Associaton (change in 1 causes change in other)

- Time Order (cause 1st, effect 2nd)

- Make sure not some other cause didn't acct for

Measures of Assoc give us an indication of the ___ of assoc & (if ordinal) the ___ of assoc between variables

Direction

Strength

Strength

Identify out of 5 scores least strong to most strong

- -.9 is stronger than +.1

- Use Absolute distance from 0.

- CAN NEVER PASS 1 (1.1 is impossible

A negative value of Gamma necessarily indicates the variables are negatively assoc.

False

Because 2 variables are strongly related they are necessarily causally related?

False

How to determine strength

.0 - .2 Weak

.21 - .49 Moderate

.5 - .69 Strong

.7 - 1 Very Strong

.21 - .49 Moderate

.5 - .69 Strong

.7 - 1 Very Strong

Scatterplot (Scattergram)

Graphs the cases at hand in terms of 2 variables. A regression line is the attempt to summarize the distribution of points in a scattergram.

Chi-Square

A test of statistical significance appropriate for nominal & ordinal variables

Degrees of Freedom

Way to force us to be a little more conservative as we estimate who we're comparing to each other weather # of categories or people

Sampling Error

The extent to which a probability sample of a given design is expected to vary from perfectly representing the population from which it was drawn

Null Hypothesis

States no relationship between variables, that the IV had no effect on the DV. Rejecting it saying they are dependent that one causes the effect of the other

Research Hypothesis

There is a relationship between 2 variables

Logic behind Rejecting the Null

The probability that you got the score you did on a statistical test is so low, less than 5%, that you're more than 95% sure the result did not happen by chance. Therefore, there is some real relationship between variables considering.

Statistical Significance Vs. Meaningfulness

Just because something is statistically significant doesn't mean its meaningful. Need to look at it societal. Also sometimes finding no statistical significance between variables can be meaningful as in disproving a myth.

How & when to use a scatterplot to chk possible correlation and how to read it

Plot on X Y axis paired data pts from scale variables. Looking at where the dots are clustered to see if it indicates an association

How to interpret strength & direction of an Association

Strength is the absolute distance from 0 with 0 being weakest & 1 being strongest.

Direction can only be interpreted if variables are ordinal. If both increase direct, if one increases & other decreases indirect.

Direction can only be interpreted if variables are ordinal. If both increase direct, if one increases & other decreases indirect.

How to calculate df for Chi-Square

df = (r-1)(c-1)

When to choose Phi, Lambda, Gamma & Pearsons r as appropriate measures of assoc.

- Phi - 2 dichotomous variables (nominal or ordinal).
- Lambda - Nominal Variables.
- Gamma - Ordinal Variables
- Pearsons r - 2 Interval/Ratio (scale) Variables

How to interpret SPSS output for Chi-Square

Describe test value, give df, what sign lve is & if reject or accept null.

EX: Chi test for (name variables) give test result of 16.007 w/1 df which is significant to p<.001. Therefore I would reject the Null

EX: Chi test for (name variables) give test result of 16.007 w/1 df which is significant to p<.001. Therefore I would reject the Null

How to interpret SPSS output for Measure of Assoc

Name the measure (phi, gamma, ect), name the variables used, write # state strength of assoc.

EX: The value for Phi has meas of Assoc between (name of variables) is .091 this indicates a weak assoc.

EX: The value for Phi has meas of Assoc between (name of variables) is .091 this indicates a weak assoc.

How to interpret SPSS output for Cross Tabulation

know IV is column DV is Row, column % calculate down & compare across. That if there is a 10% or diff worth talking about

Univariate Analysis

1 variable

Major Question - what

Primary Goal - Description

Major Question - what

Primary Goal - Description

"The semester I found StudyBlue, I went from a 2.8 to a 3.8, and graduated with honors!"

Jennifer Colorado School of Mines
StudyBlue is not sponsored or endorsed by any college, university, or instructor.

© 2014 StudyBlue Inc. All rights reserved.

© 2014 StudyBlue Inc. All rights reserved.