aStat 0.6

What's new?

• Confidence intervals estimation for a single proportion
• Three methods are used for constructing confidence intervals: exact method, Wilson's method, and Agresti-Coull method

aStat 0.5

What's new?

• Number needed to treat or harm calculation
• The confidence interval is constructed using Newcombe's method.
• Fixed: monetary calculation is used for person-time analysis

aStat 0.4

What's new?

Person-time Data Analysis

• Test of significance: Chi-square p and exact p values
• Incidence rate ratio and confidence intervals (CIs) constructed using binomial exact method and Graham's score-based method
• Incidence rate difference and CI estimated approximately

Other minor changes

• Demo data added
• BigDecimal (Java) was used in some calculation although this may be not necessary.

aStat 0.3

Minor change: using JSci library instead of java commons math library, to make the file smaller after  installation. (~ 700k)

aStat 0.2

Version 0.2

New features:

Risk difference between the exposed group and the unexposed. Confidence interval is constructed using the Newcombe's method.

aStat 0.1

aStat is an easy-to-use, user-friendly statistical calculator for scientists.

It is not intended to substitute with those professional statistical packages like SPSS. Instead, the aim is to provide a handy calculator for scientists and researchers when appropriate. Unlike other statistical programs in the android market, aStat does not provide statistical distributions or complex functions like multivariate analysis. In particular, it offers robust yet easy-to-understand confidence interval estimates in 2x2 contingency table analysis, which is an important pre-analysis before ones step into further multivariate regression.

Current features of 2x2 contingency analysis include:

• Inference of significance: Chi-square test, Yate's Chi-square test, and Fisher's exact test (by summation of all small p values)
• Inference of association: odds ratio and risk ratio with confidence intervals, using Wald-based approximation and the Cox-Hinkley-Miettinen-Nurminen method.
• Calculation of diagnostic power: positive predictive value (PPV), negative predictive value (NPV), sensitivity, and specificity.
• Confidence interval (CI) for smalls counts: Agresti-Coull method and Wilson's method are used for 95% CI estimation. Both are more suitable for small counts (< 40-50) than usual Wald-based approximation
• Settings for non-zero decimals and confidence level: decimals: 2-6; confidence level: 90%, 95%, 99%.

Features to be implemented:

• Person-time inference using exact test
• Confidence intervals for counts or proportions
• Chi-square test for trend in binomial proportions