Statistical test for Non continuous variables. Dr L.M.M. Nunn
What does the term “statistics” mean? A statistic is an estimate, based on random sampling of the population, of parameters of the population. Emphasis on statistical analysis in research P < 0.05 Statistically significant P > 0.05 Statistically insignificant Statistical testing > individual data points
Probability: Numerical likelihood of the occurrence of an event. Significant: p < 0.05 Why 5% as level of statistical significance? If p < 0.05, it means that the likelihood that the event was due to chance is < 5%. Thus > 95% certainty that the event was not due to chance.
Hypothesis testing: Likely or unlikely to occur. Convert question into Null hypothesis H0 = No difference between sample + population. H1 = Alternate hypothesis = what you are trying to prove
Hypothesis testing (cont.) Example : Aspirin vs placebo in MI patients H0: aspirin = placebo H1: Aspirin > placebo If α < 0.05: reject null hypothesis and accept H1. i.e. Aspirin more advantageous than placebo in MI patients.
Variables: Ordinal: Ordered Relative rather than absolute relations btw variables: eg: Apgar scores Power (1- 5) Level of pain (0 – 10)
Nominal variables: Named Quality rather than quantity eg. Female + Male Alive + dead EEG waveforms (α, β, θ, δ)
Quantitative Variables: A. Discrete: Limited no of possible variables eg. No. of previous pregnancies No. of cases of acute cholecystitis B. Continuous variables Unlimited no of possible variables eg. height, weight
Selecting appropriate statistical test: 1. Nominal : Chi square test Fisher exact test 2. Ordinal : Parametric (Normal distribution, large sample size) Non parametric test (Abnormal distribution small sample size).
3.Continuous variables: Analysis of linear regression.
Contingency tables: Ordinal & nominal scales different techniques available for presentation + analysis of results Histograms are of limited value Nominal data: Chi square test best Contingency table No. of rows and columns eg, 2x4
2x2 Contingency table A B + _
Chi Square test: x²= sum of (observed – expected no. of individuals in a cell)² / expected no. of individuals in a cell. x² = Sum of (0 – E)² E
Observed frequencies similar to expected frequencies then x² = small no. i.e. statistical insignificant. Observed + expected frequencies differ then X² = big no. and statistically insignificant
Chi Test (continued): Test whether data has any given distribution Frequency table yielding observed frequencies. Probabilities calculated for each category Probabilities converted into frequencies = expected frequencies Compare observed frequencies with expected frequencies.
Observed frequencies similar to expected frequencies, then the observed frequency distribution is well approximated by hypothesis one.
Fisher Exact Test: The Chi square test used to analyze 2x2 contingency tables when frequency of observations in all cells are at least 5 In small studies when expected frequency is <5: Fisher Exact test Turns liability of small sample sizes into a benefit.
Sensitivity: Proportion of cases correctly diagnosed by a test = sensitivity or Sensitivity of a test is the probability that it will correctly diagnose a case Screening test eg. Rapid HIV
Specificity: Proportion of non cases correctly classified by a test. Or Specificity represents the probability that a non case will be correctly classified If a +ve test results lead to major intervention eg, colectomy, mastectomy, a high specificity is essential. Test lacks specificity a substantial no. of people may receive unnecessary & injurious treatment.
Predictive value: Predictive value of a test depends on the prevalence of disease in the population of patients to whom it is applied.
Disease Test TP FP - FN TN
Sensitivity = TP (TP + FN) Specificity = TN (TN + FP) Positive predictive value = TP (TP + FP) Negative predictive value = TN (FN + TN)
Summary Statistical tests provide the investigator with a “p” value. Choose the correct Statistical test according to the appropriate Variable. “p” value < 0.05, Statistically significant,Null hypothesis is rejected and Alternate hypothesis accepted.