Reliability and Validity of Researcher-Made Surveys.

Reliability and Validity of Researcher-Made Surveys

Reliability

Reliability

“Authors should provide

reliability coefficients of

the scores for the data

they analyze even when

the focus of their

research is not

psychometric.”

Reliability

“Authors should provide

reliability coefficients of

the scores for the data

they analyze even when

the focus of their

research is not

psychometric.”

Reliability is the amount

of random fluctuation in

individual scores.

Reliability

“Authors should provide

reliability coefficients of

the scores for the data

they analyze even when

the focus of their

research is not

psychometric.”

Reliability is the amount

of random fluctuation in

individual scores.

“In practice, score

reliability is a matter of

degree, because all

scores include some

random fluctuation.”

Reliability

“It is the reliability of the data in hand in a given study that will drive study results, not the reliability of the scores described in the test manual.”

Reliability is the amount

of random fluctuation in

individual scores.

“In practice, score

reliability is a matter of

degree, because all

scores include some

random fluctuation.”

Reliability

“It is the reliability of the data in hand in a given study that will drive study results, not the reliability of the scores described in the test manual.”

“Poor score reliability

may compromise …

score ‘validity’.”

“In practice, score

reliability is a matter of

degree, because all

scores include some

random fluctuation.”

Reliability

“It is the reliability of the data in hand in a given study that will drive study results, not the reliability of the scores described in the test manual.”

“Poor score reliability

may compromise …

score ‘validity’.”

“Poor score reliability

may compromise the

ability of a study to yield

noteworthy effects”

Reliability

“Poor score reliability

may compromise …

score ‘validity’.”

“Poor score reliability

may compromise the

ability of a study to yield

noteworthy effects”

Reliability

“Scores can’t both measure nothing and measure something.”

“Poor score reliability

may compromise …

score ‘validity’.”

“Poor score reliability

may compromise the

ability of a study to yield

noteworthy effects”

Reliability

“Scores can’t both measure nothing and measure something.”

“Perfectly unreliable scores are perfectly random and cannot yield… significant results.”

“Poor score reliability

may compromise …

score ‘validity’.”

“Poor score reliability

may compromise the

ability of a study to yield

noteworthy effects”

Reporting Reliability

“Scores can’t both measure nothing and measure something.”

“Perfectly unreliable scores are perfectly random and cannot yield… significant results.”

“Poor score reliability

may compromise …

score ‘validity’.”

“Poor score reliability

may compromise the

ability of a study to yield

noteworthy effects”

Reporting Reliability

Reporting Reliability

“Reporting reliability coefficients for one’s own data is the exception rather than the norm...Too few reliability estimates for analyzed data are provided in both journals…and doctoral dissertations.”

Reporting Reliability

“Reporting reliability coefficients for Reporting reliability coefficients for one’s own data is the exception one’s own data is the exception rather than the norm...rather than the norm...Too few Too few reliability estimatesreliability estimates for analyzed for analyzed data data are provided in both journals… both journals…and and doctoral dissertations.”

Reporting Reliability

“The most commonly used (reliability) estimate is Cronbach’s (1951) coefficient alpha ().”

“The most commonly used (reliability) estimate is Cronbach’s (1951) coefficient alpha ().”

number of items (sum of item variances)

number of items-1 test variance

Reporting Reliability

1-

“The most commonly used (reliability) estimate is Cronbach’s (1951) coefficient alpha ().”

“Item score covariance plays an important role in both the numerator and the

denominator of the estimate.”

Reporting Reliability

“The most commonly used (reliability) estimate is Cronbach’s (1951) coefficient alpha ().”

“The intercorrelations of the items are the essential source of this

kind of reliability.”

Reporting Reliability

Validity of Researcher-Made Surveys

Evidence of Validity

Evidence of Validity

• Patterns of Association

Evidence of Validity

• Patterns of Association

• Comparing Results from Different Versions of the Same Question

Evidence of Validity

• Patterns of Association

• Comparing Results from Different Versions of the Same Question

• Comparing Responses to Data from Other Sources

Evidence of Validity

• Patterns of Association

• Comparing Results from Different Versions of the Same Question

• Comparing Responses to Data from Other Sources

• Asking the Same Question Twice and Comparing Results

• Patterns of Association

• Comparing Results from Different Versions of the Same Question

• Comparing Responses to Data from Other Sources

• Asking the Same Question Twice and Comparing Results

Evidence of Validity

Reliability

Reliability

• Patterns of Association

Evidence of Validity

• Patterns of Association• Scores from different measures believed to measure

similar things should correlate. Scores from different measures believed not to measure similar things should not correlate.

• Responses to items believed to represent the same dimensions or factors should correlate.

Evidence of Validity

• Comparing Responses to Data from Other Sources

Evidence of Validity

• Comparing Responses to Data from Other Sources

• Compare to records. Compare to physical testing. Compare to population estimates.

Evidence of Validity

Face Validity of Survey Questions

Face Validity of Survey Questions

• Have a reason for every question you ask.

• Keep questions simple.

• Keep questions precise.

• Avoid leading questions.

• Foresee social desirability.

• Response options should be mutually exclusive and exhaustive.

• Provide temporal frame of reference.

• Use Likert format correctly.

Pilot Testing

Pilot Testing

• Sample size >15

• Discuss questions with respondents to find confusion or ambiguity.

• Pretest sample should resemble study sample.

• Examine variance among respondents.

• Refine answer options.

• Time how long it takes.

Group Assignment

Produce a Valid and Reliable Attitude or Psychological Scale in 90 Minutes

Produce a Valid and Reliable Attitude or Psychological Scale in 90 Minutes

Write a 7 to 10 item scale. 30 minutes

Pilot test your items. 15 minutes Use another group. If necessary, revise your scale based on your pilot testing.

Administer the revised scale to 10 minutes at least 6 people. Collect data.

Enter data on SPSS. 20 minutes Compute coefficient alpha. Revise your scale based on reliability data.

Report. 15 minutes

Entering Survey Data on SPSSItem1 Item2 Item3

Survey1 4 1 5

Survey2 5 1 4

Survey3 4 2 3

Item 1 “I like salt.”

Strongly Disagree Disagree Neutral Agree Strongly Agree

1 2 3 4 5

Analyzing for Reliability in SPSS1. Enter survey data.2. Choose menu options:Analyze

ScaleReliability Analysis

3. Choose and move variables to Items box.4. Click Statistics…Click on: Item

ScaleScale if Item Deleted

5. Read output. Identify items which have an “Alpha if Item Deleted” larger than the scale’s Alpha. Remove those items.

6. Re-run analysis with remaining items until satisfied.