Login to use UniMult now!

Login or Create a new Account
 
     
Forgot Login?   Sign up  

Statistical Analyses

Easier for the occasional user:

Just 4 simple steps.

1. Enter the names of your variables and the data for each case.

2. Select your Y's (dependent variable, outcome).

3. Select your X's (independent variables, predictors)

4. Examine your results.

UniMult labels the results appropriately, whether cross-tab, ANOVA, or regression analysis.

No need for you to name a statistic / analysis (the Guide explains why)

 

Sophisticated for the seasoned user:

UniMult does both univariate and multivarate data analyses (hence the name).

Uses a UNIfied MULTivariate GLS model of which all standard analyses are special cases. (The Guide explains the simple generalization to multivariate analyses.)

Mix dichotomous, nominal, and other variables in any order as Y's or X's, including interactions between all types of variables.

Confidence intervals, effect sizes, and odds ratios.

Checks for univariate outliers with every data set.

Simple power analysis given for every data set.

Partial out any variable(s) -- even nominal -- from Ys or Xs or both.

Gives multivariate analysis when you select more than one Y. This increases df2, resulting in more power to detect results with moderate sample sizes.

Easily run family-wide tests to protect overall Type 1 error rates.

Split overlapping variances by sequential / hierarchical partialling, orthogonalization, or factor / component analysis.

Factor / component analysis includes higher order factors. Factors / components become variables and can be immediately included in statistical analysis with non-factored variables in the data set.

Classical test analysis using a better procedure than item-total or item-remainder correlations.

Spend your time identifying questions to ask and interpreting the results (leave the rest up to your statistical servant, UniMult).

Some advanced data analysis techniques are not currently scheduled to beincluded, generally because good programs already exist (e.g., SEM, IRT) or lack of useful applications (e.g., canonical variates).