Welcome to Stat4all (beta)


Stat4all (beta) is a web-based statistical package that automates the selection of quantitative data analysis procedures and the generation of ad hoc statistical analysis reports. Our goal is to make Stat4all a suitable statistical analysis package for all users, regardless of their level of experience in statistics.

As of April 2018, the open beta version of Stat4all supports various basic statistical procedures, from comparisons (e.g., t-test, ANOVA...) to building linear regression models. We will continue adding new features to include more advanced statistical methods, as well as settings to enable our advanced users more control over the parameters and the outputs.

Get Started

To get started using Stat4all you can either check the Help section or start clicking around to see what happens. Don't worry if you break something, that's what a beta test is for. Just make sure to visit the Contact section and notify us when that happens.

Before you begin with the analysis, you'll have to upload a dataset in the section Data as we don't support importing data from other services at the moment. If you need to change column data types or create composite variables before running an analysis, visit the Transform section. We also recommend you check the Settings to make sure you agree with the configuration of certain parameters like the desired significance level. Once you're set up, click on the Analyze section in the side bar to list the available procedures.

Data Set Preview


Compare groups


Pay What You Want (optional)

As most parts of our app are still in development or in beta, it is free to use at the moment, no registration required. However, if you like our statistical package and want to support its development, you can pay what you want for using it. All contributions, one-time and recurring, are appreciated!


One-time Payment Option

If you want to make a one-time payment, you can enter any amount you want in USD when you click on the button below.


Monthly Payment Options

Examine Relationships between Variables


Pay What You Want (optional)

As most parts of our app are still in development or in beta, it is free to use at the moment, no registration required. However, if you like our statistical package and want to support its development, you can pay what you want for using it. All contributions, one-time and recurring, are appreciated!


One-time Payment Option

If you want to make a one-time payment, you can enter any amount you want in USD when you click on the button below.


Monthly Payment Options

Important

The term linear model in statistics is not used to describe the trend between two variables. The model is linear with respect to the parameters, i.e., your independent/predictor variables, not the dependent/criterion variable. A linear model has a form of a function that satisfies the following:

  1. Each independent/predictor variable is multiplied by one unknown coefficient;
  2. There can be only one unknown parameter in the model without an associated variable, i.e., the constant;
  3. The output of the function is the sum of all individual terms.

Therefore, even if you perform a polynomial regression, or any other type of transformation of the independent/predictor variables, it's a linear model as long as the function doesn't violate the requirements listed above.

Pay What You Want (optional)

As most parts of our app are still in development or in beta, it is free to use at the moment, no registration required. However, if you like our statistical package and want to support its development, you can pay what you want for using it. All contributions, one-time and recurring, are appreciated!


One-time Payment Option

If you want to make a one-time payment, you can enter any amount you want in USD when you click on the button below.


Monthly Payment Options

Pay What You Want (optional)

As most parts of our app are still in development or in beta, it is free to use at the moment, no registration required. However, if you like our statistical package and want to support its development, you can pay what you want for using it. All contributions, one-time and recurring, are appreciated!


One-time Payment Option

If you want to make a one-time payment, you can enter any amount you want in USD when you click on the button below.


Monthly Payment Options

Most journals and assignments use p < 0.05 as the cut-off p-value, but you can change the desired p-value cut-off to 0.01 if you need to be more conservative or to 0.10 if are allowed to make a less conservative interpretation of your results.


Stat4all uses the Shapiro-Wilks test by default to test the assumption of normality for parametric tests, but you can also choose to use the Kolmogorov-Smirnov or skewness and kurtosis instead. Both the S-W and K-S tests can be sensitive to departures from normality as sample size increases, so using skewness and kurtosis for large samples is recommended. Both valuesneed to be in the range 0.00 ± 2.00 to assume normality (Field, 2000; Gravetter & Wallnau, 2014; Trochim & Donnelly, 2006).


As a general rule, parametric tests become more robust to departures from normality and other assumptions as the sample size increases, starting at N of 30-40. However, we recommend running parametric tests with non-normal data only when N > 200 (Fagerland, 2012).

How to Use Stat4all

1. Upload or Import* Data

The beta version of Stat4all doesn't offer permanent file storage, so the data you upload will be available only for as long as your session remains active. We support the following file formats:

  1. Separated data (.csv, .psv, .tsv)
  2. SAS files (.sas7bdat, .xpt)
  3. SPSS files (.sav, .por)
  4. Office files (.xlsx, .xls, .ods)
  5. Stata files (.dta)
  6. Minitab files (.mtp)

Although Stat4all supports various file formats, the data should be organized as a standard data matrix, which means that:

  • Each row of the data matrix represents one observation.
  • Each column of the data matrix represents one variable.

The upload limit is one 5 MB file, which may not sound like a lot, but a 5 MB file can hold thousands of observations with hundreds of variables. (Obviously, that is not true for all data types and formats, but it's true for most quantitative data sets.)

If you don't have a data set, but would like to see how the application works, you can select the Example Datasets tab and choose from various datasets available in R's datasets package.

* Importing data from third-party sources is not available in the beta version.


2. Settings

In most cases you won't need to change the default settings, but you can change (a) the critical p-value for rejecting the null hypothesis and accepting the research hypothesis, (b) the test used to test the normality assumption, and (c) the preference for parametric tests in large samples. The rationale behind the default settings and situations in which you could change them are explained in the Settings section.


3. Transform Data

When you upload your data set, the application stores the variables as numbers (integer and double data types) or characters. In the Transform section you can check the current data types and transform variables from one type to another as necessary.

3.1. To characters and numeric

Characters and numeric data types are self-explanatory, but here are some things to keep in mind when changing variable data types into one of those two:

  • The application won't let you transform factors to numeric because a factor is an integer with a label. If you transform it into a numeric factor, it will discard the labels and show you a set of integers, so the package allows you to turn factors into ordered factors or characters only.
  • When you transform a character into a numeric variable, it will coerce anything that isn't a number into a missing value. For example, if you have the value NULL in the original data set, it will be coerced into a missing value during the transformation.
  • The current version of Stat4all recognizes only the period (.) as the decimal place separator, so it won't transform numbers with a comma (,) decimal separator into a numeric vector.

3.2. To factors and ordered factors

Factors are used to store categorical variables in R and are usually used as grouping variables.

Most analyses will transform a character or numeric into a factor type, so you don't have to, but in some cases you'll want to do it if you want the output to make sense. For example, if you code values "Control" and "Treatment" as 0 and 1 in the data entry process, the software will recognize those values as numeric. Even though it will change the numeric type into a factor, it will use "0" and "1" as labels to refer to your groups in the output. To prevent those scenarios, you can specify the labels when changing numeric types into factors.

Ordered factors should be used when you plan on running a regression analysis with an ordinal variable as one of the predictor/independent variables. You have to list the labels in ascending order when transforming a variable into an ordered factor. Stat4all will refuse to perform the transformation if you don't provide the labels because it doesn't know how to order the labels.


4. Analyze

The beta version allows you to run analyses that compare groups (Analyze > Compare Groups), find correlation coefficients (Analyze > Examine Relationships), compare repeated measures (Analyze > Repeated Measures), and build simple or multiple linear regression models (Analyze > Regression Models). In future releases we plan on including more statistical techniques, but the ones we currently support are sufficient in most cases.

Every analysis, with the exception of a regression analysis, test assumptions such as normality and homogeneity of variance to choose tests. The regression analysis produces the model first, after which it provides you with information about the model's assumptions.

4.1. Comparing Groups

In the Analyze > Compare Groups section, select the dependent variable in the first drop-down menu and the grouping variable in the second one. You can select two grouping variables, in which case you can choose to include the interaction between the two as .

The procedure will first check the number of participants in each group. It won't be able to analyze the data if one or more groups lack the minimum number of participants to produce an accurate statistic and p-value.

Depending on the results of the descriptive statistics and your Settings, the procedure will use either a t-test/Wilcoxon or ANOVA/Kruskal-Wallis with post hoc tests if you provide one grouping variable or a two-way ANOVA/Scheirer-Ray-Hare rank sum test with post hoc tests.

4.2. Repeated Measures

In the Analyze > Repeated Measures section you'll find two tabs, but only the pre-/post test is available in beta. When you select the pre-test values and post-test, the report will provide y

4.3. Examine Relationships

The word relationships in Analyze > Examine Relationships refers to relationships between two or more variables. Once you select the variables you want to include in a bivariate correlation analysis, you can also select which of those variables you want as covariates in the analysis. Perforing a partial correlation is optional, if you leave that field blank, you'll just get the results of the bivariate correlations.

You can assume a linear or non-linear relationship between the variables. Stat4all will use Pearson's method if you assume linearity and Spearman's if you assume non-linearity. The default is linear, but you'll get a multivariate outlier analysis in the appendix of the report, so you can rerun the analysis if you find outliers that can influence the linearity assumption.

4.4. Regression Models

Simple and multiple linear models are supported in beta.

In the Simple tab, select the dependent or criterion variable in the first drop-down menu and the independent or predictor variable in the second drop-down menu. The default transformation of the independent or predictor variable is "None", but you can choose between "Log", "Quadratic", and "Cubic" as well if the relationship between the two variables is not linear. (It's still a linear model even if you transform the independent variable because the term linear refers to the parameters, not the shape of the trend line between the two variables.)

In the Multiple tab select you'll see the same options, without the transformation of independent variables, but you can select multiple independent variables. Building models that include interactions between independent variables is not possible at the moment.


5. Contact

Our contact information is listed in the Contact section. If you have questions, recommendations, or want to report problems you encounter while using Stat4all, we'd like to hear from you! Just remember that we don't provide technical or any other kind of support because the software is still in beta and free to use.

Contact Us

Company Information

Stat4all is a trademark of White Dog Analytics J.D.O.O., registered at the Commercial Court in Zagreb, Croatia, MBS: 081127850, OIB: 89817123449, VAT ID: HR89817123449

Headquarters

WHITE DOG ANALYTICS J.D.O.O.
Hecimoviceva ulica 1
10000 Zagreb
Croatia / Hrvatska
E-mail: info@whitedoganalytics.com
Social: Facebook | Twitter | LinkedIn | YouTube

Please keep in mind that we don't offer technical support for beta software, which is provided 'as is' according to our Stat4all BETA License Agreement However, If you encounter a problem while using our statistical package or if you would like to suggest new features or improvements to existing features, please feel free to contact us.