Créer une présentation
Télécharger la présentation

Télécharger la présentation
## PASW-SPSS STATISTICS

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -

**PASW-SPSS STATISTICS**• David P. Yens, Ph.D. • New York College of Osteopathic Medicine, NYIT dyens@nyit.edu • PRESENTATION 3 • Descriptive Statistics • Chi-Squared • Risk/Odds Ratio 2010 David Yens, Ph.D. NYCOM**DESCRIPTIVE STATISTICS**• When doing data analyses, you usually want to see the nature of the data before you start. You get this from • FREQUENCIES for nonparametric data and • DESCRIPTIVES for parametric data**FREQUENCIES**• ANALYZE • DESCRIPTIVE STATISTICS • DESCRIPTIVES • You have data on length of stay for a large sample of patients and want to examine the parameters for age and length of stay.**FREQUENCIES**• ANALYZE • DESCRIPTIVE STATISTICS • FREQUENCIES • In your length of stay data you have included information about gender. How many males and females are in the data?**JOINT FREQUENCIES**• The next question might be whether there is a difference in the number of admissions by gender.**CATEGORICAL FREQUENCY DATA: TESTS OF SIGNIFICANCE**• CHI-SQUARED (χ2 ) • Contingency table • Test of association; compares proportions • Assesses signal-to-noise ratio • Based on the differences between observed and values and expected values • Most often used with 2 x 2 tables • Yates’ correction • Fisher’s exact test David Yens, Ph.D. NYCOM**THE RELATION BETWEEN OBSERVED AND EXPECTED FREQUENCIES**• if the null hypothesis is true, the absolute value of the differences between the observed and expected cell frequencies will, on balance, be small; • if the null hypothesis is false and the alternate hypothesis is true, the absolute value of the differences between the observed and expected cell frequencies will, on balance, be large. David Yens, Ph.D. NYCOM**CHI SQUARED**• The test statistic is given by • χ 2 = ∑ ( O – e)2/ e David Yens, Ph.D. NYCOM**CATEGORICAL FREQUENCY DATA: TESTS OF SIGNIFICANCE**• CHI-SQUARED 2x2 • A table in which frequencies correspond to two variables. (One variable is used to categorize rows, and a second variable is used to categorize columns.) • Contingency tables have at least two rows and at least two columns. • Test of association; compares frequencies • Based on the differences between observed and values and expected values • Most often used with 2 x 2 tables David Yens, Ph.D. NYCOM**2x2 CHI-SQUARED**• First, we create a 2x2 contingency table, as shown below. Assume that in the treatment group 15 subjects had a positive response and 10 and a negative response, and for the control group 5 subjects had a positive response and 20 had a negative response. The letters on the table at the left identify the letters used in the formula below; the sample data table is on the right. • For a 2x2 table, the critical value is 3.84. If the Chi-Squared you calculate is > 3.84, the result is significant at p<.05. David Yens, Ph.D. NYCOM**SPSS CROSSTABULATION**• ANALYZE • DESCRIPTIVE STATISTICS • CROSSTABS • Note that for a Chi-Squared analysis an expected cell frequency of 5 or more is preferred. If less than 5, use Fisher’s Exact Test or Yates’ correction**Yates’ Correction for Small Numbers**• Used if expected frequency for a cell is <5 χ 2 = Σ [|Oi – Ei|-.5]2/Ei David Yens, Ph.D. NYCOM**Fisher’s Exact Test**• For full computation for values as extreme or more extreme than the one observed, must compute the probability for each extreme case and sum the probabilities • Fisher’s Exact Test – for a 2x2 analysis with small numbers in each cell: David Yens, Ph.D. NYCOM**PROBLEM**• Using a database of toothbrushing activity by children, we would like to know whether there is a difference between brushing activity by boys and girls. The data contain gender and whether or not they brush daily. • These are frequency data and appropriate for crosstabs with a Chi-Squared statistic. • (See Chapt. 7 of IBM SPSS)**DATA LAYOUT**• GenderDailyBrushing • M Y • M N • M N • M N • M Y • M N • M N • F Y • F Y • F Y • F Y • F N • F Y • F Y • F Y**CROSSTABULATION**• ANALYZE • DESCRIPTIVE STATISTICS • CROSSTABS • Crosstabs provides access to other analyses: • Risk Ratios and Odds Ratios (pp. 114-116) • Relative Risk: The ratio of incidence in exposed (or group) of persons to incidence in nonexposed (other group) persons • Odds Ratio – The odds that a case is exposed divided by the odds that a control is exposed**RELATIVE RISK**• RELATIVE RISK (Cohort studies) Ratio of the risk of disease in exposed individuals to the risk of disease in nonexposed individuals Relative Risk = = David P. Yens, Ph.D. NYCOM**ODDS RATIO**• ODDS RATIO (Cohort studies) Ratio of the odds of development of disease in exposed individuals to the odds of development of the disease in nonexposed individuals Odds Ratio = David P. Yens, Ph.D. NYCOM**PROBLEM**Consider the data taken from a study that attempts to determine whether the use of electronic fetal monitoring (EFM) during labor affects the frequency of cesarean section deliveries. The 5824 infants included in the study, 2850 were electronically monitored and 2974 were not. The outcomes are as follows: • Calculate the odds ratio associated with EFM exposure.**SOLUTIONFor this analysis, the raw data are reduced to a 2**by 2 table with Crosstabs and then subsequently analyzed by hand**CROSSTABULATION**• ANALYZE • DESCRIPTIVE STATISTICS • CROSSTABS • Crosstabs provides access to other analyses: • Kappa – provides measure of agreement between 2 judges: Cohen's kappa measures the agreement between the evaluations of two raters when both are rating the same object. A value of 1 indicates perfect agreement. A value of 0 indicates that agreement is no better than chance. Kappa is available only for tables in which both variables use the same category values and both variables have the same number of categories.**CROSSTABULATION**• ANALYZE • DESCRIPTIVE STATISTICS • CROSSTABS • Crosstabs provides access to other analyses: • The 2 by 2 tables also provide the basis for several other epidemiological computations**PROPORTIONS/PERCENTAGES**The relationship between prior condom use and tubal pregnancy was assessed in a population-based case-controlled study at Group Health Cooperative of Puget Sound during 1981-1986. The results are: Compute the proportion of subjects in each group who never used condoms.**SENSITIVITY**• SENSITIVITY - • Accuracy of the test in detecting the condition in patients who actually have it • Sensitivity Se = David P. Yens, Ph.D. NYCOM**SPECIFICITY**• SPECIFICITY - • How well the test correctly identifies patients who do not have the condition • Specificity Sp = David P. Yens, Ph.D. NYCOM**PROBLEM**Consider the following data: Calculate the sensitivity and specificity of X-ray as a screening test for tuberculosis. SOLUTION: • SENSITIVITY = 22/30 = .73 • SPECIFICITY = 1739/1790 = .97**EPIDEMIOLOGY**• INCIDENCE - EXPOSED • Number of new cases of a disease that occur during a specified period of time in a population at risk for developing the disease Incidence in exposed = David P. Yens, Ph.D. NYCOM**EPIDEMIOLOGY**• INCIDENCE - NONEXPOSED • Number of new cases of a disease that occur during a specified period of time in a population at risk for developing the disease Incidence in Nonexposed = David P. Yens, Ph.D. NYCOM**EPIDEMIOLOGY**• PREVALENCE - • Proportion of patients in a given population who have a given disease • Prevalence, P = David P. Yens, Ph.D. NYCOM**EPIDEMIOLOGY**• LIKELIHOOD RATIO - • The odds that a test result occurs in patients with the disease versus those without the disease Positive Likelihood Ratio, LR+ = ----------------- David P. Yens, Ph.D. NYCOM