Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.


Need some help? Watch this video.

Caution!

This software is in development phase. It may have some bugs and its design and features may be subject to change.

Get started

To get to the next step of the analysis hit the next button in the upper part of the sidebar. For a short introduction you can also watch this video.

Welcome to multigrid one

multgrid one is a software for analyzing multiple grids using mixed models. The app is designed to guide you through the process of preparing the data, defininig the model and interpreting the results. It is not necessary to have expert data analysis skills to use it. The goal of the software is to facilitate the application of mixed model analysis and repertory grid data for casual users.

Steps of analyis

  1. 0.

    Read the info page.

  2. 1.

    Upload your grid data and additional information for each grid, if there is some.

  3. 2.

    Explore your grids. This is an optional step but may help to make sure the data was read in as you expected.

  4. 3.

    Select the grid indexes you want to calculate for the analysis.

  5. 4.

    Let the computer prepare your data. You may also download the prepared data to use it elsewhere.

  6. 5.

    Define and run the model you want to estimate and explore the results.

More infos and troubleshooting

The best browser to display this app is Google Chrome. If you experience troubles with the display, give it a try.

If you would like to report a bug or have questions/suggestions for multigrid one, please visit our discussion board.

Developers

Mark Heckmann, Richard C. Bell

Load grids and external data

On the tabs below you can upload grid files. If you want to include additional information for each grid into the analysis you may also upload additional data. The second step is optional.

Upload grids

To get started you need to upload some grids. For more information on the data formats that are supported click on the More Info link on the lower right.

Get some data In case you do not have any grids available you can download sample grid data here (zip file).


Format

You can upload grid data that comes in different formats. Currently the following formats are supported:

  • Standard text files

In the future the following file types will be supported:

  • Gridcor
  • Gridstat
  • GridSuite
  • sci:vesco
  • Excel

.txt Format

You can define a grid using a standard text editor and saveing it as a .txt file. The content of the .txt file has to be in a special format. There are three mandatory blocks each starting and ending with a predefined tag in uppercase letters. The first block starts with ELEMENTS and ends with END ELEMENTS and contains one element in each line. The other mandatory blocks contain the CONSTRUCTS, RATINGS and RANGE (see below). In the block containing the CONSTRUCTS the left and right pole are seperated by a colon (:). The order of the blocks is arbitrary. All text not contained within the blocks is discarded and can thus be used for comments.

------------ example.txt file ---------------

anything not contained within the tags will be discarded

ELEMENTS
element 1
element 2
element 3
END ELEMENTS

CONSTRUCTS
left pole 1 : right pole 1
left pole 2 : right pole 2
left pole 3 : right pole 3
left pole 4 : right pole 4
END CONSTRUCTS

RATINGS
1 3 2
4 1 1
1 4 4
3 1 1
END RATINGS

RANGE
1 4
END RANGE
---------------- end of file ----------------


You can include additional information on the grid level (e.g. depression scores for each person) or on the constructs level (e.g. an importance rating for each construct) into the analysis. To do this you need to upload a datafile containing the information.

Get some data In case you do not have any additional data available you can download this datafile (csv file) that corresponds to the grids from the previous tab.

Skip step? If you do not want to include additional information in the analysis skip this step.



Format

The datafile you upload needs to be in CSV format. In order to match the external data with the calculated data, the datafile must contain the variable(s) grid (and construct).

Grid level data

To include additional information on the grid level, the file must to contain a variable named grid which is a running index for each grid. The numbering of the grids corresponds to the alphabetic order of the grid filenames you have uploaded. Below, a sample CSV file for two grids (grid) and the scores for the variables depression and anxiety is shown.

"grid","depression","anxiety"
1,21,12
2,12,19

Construct level data

To include external data on the construct level the datset must additionally contain the variable construct which is a running index of the construct number. The other variables contain the information you want to include. Below, a sample CSV file for two grids (grid) with three constructs each (construct) and the scores for the variables importance (construct level) and depression and anxiety (grid level) is shown.

"grid", "construct", "importance","depression","anxiety"
1,1,10,21,12
1,2,7,21,12
1,3,5,21,12
2,1,9,12,19
2,2,10,12,19
2,3,6,12,19

Explore some single grids

Here you have a chance to look at single grids before the multigrid analysis starts. The more comprehensive software for single grids is OpenRepGrid OnAir, here you only have some rudimentary options.



Select variables and indexes

For mixed models you can use input data on different hierarchical levels. Use the following tabs to select which variables, indexes and measures you want to include in the prepared dataset on the grid as well as on the construct level.

Info Uncheck all indexes you do not want in the dataset.

Important If many grids and many indexes are included, the preperation may take a while.

The following measures will be calculated for each grid (variable name in parenthesis):


The following indexes will be calculated for each construct (variable name in parenthesis):


Prepare data

On the previous page you have selected the variables and indexes you want to include into the dataset. Press the button to prompt the dataset preparation. After the data has been prepared you can proceed to the model section. Be patient in case a large number of grids are processed.

Info You can download the processed data as a .csv file. This may be useful if you want to apply other models which are not available here.
Download

Coffee time ;)

Depending on the number of grids and the number of indexes requested this process may take up to 1 minute.


Model and variables

Up to now you have created a dataset containing all the necessary information for the analysis. Below you can select between several scenarios and the corresponding model type. After choosing the appropriate model you can select which variables to use in the model.

Select a model

Note Here, only several standard scenarios are considered. If you want to specify other models you need to downlaod the dataset and do this in any stats software.

Nullmodel

The nullmodel is a model without any fixed predictors. It may serve to determine how much variance is explained due to variation of the random variable.

Example If construct level intensity scores are used as the dependent variable, the approach will tell you how much variance is due to variation within each person (grid) and variation between different persons. Essentially, you will get an intraclass correlation coefficient (ICC) as a result.

Select the variables first and then hit the run button.

Dependent variable

The dependent variable will usually be on the lowest hierarchical level (i.e. construct level).

Example We might want to understand how construct intesity scores which are nested within persons vary.


Random variable

A random variable may vary randomly across the levels of another variable. In the nullmodel the Intercept is allowed to vary across some other variable, e.g. the person.


The variable

may vary across the levels of variable


Example If we assume that each persons will differ with regard to their mean construct intensity, it is useful to let the mean intensity value (i.e. the intercept) vary randomly across persons (i.e. grids).

Differences-between-groups model

This model is like the nullmodel but you can additionally add a grouping factor to see if the levels of a fixed effect differ with regard to the dependent variable.

Example If construct level intensity scores are used as the dependent variable, you may assess if there are group differences e.g. between patients and controls.

Select the variables first and then hit the run button.

Dependent variable

The dependent variable will usually be on the lowest hierarchical level (i.e. construct level).

Example We might want to understand how construct intensity scores which are nested within persons vary as well as if they differ across two or more different groups.


Random variable

A random variable may vary randomly across the levels of another variable. For grids, usually the Intercept is allowed to vary across some other variable, e.g. the person or grid.


The variable

may vary across the levels of variable


Fixed variable

A fixed effect, e.g. a factor to distinguish the control and treatment group.


The following variable distinguishes the groups


Example We want to assess if the intensity scores on constructs level differ between a treatment and a control group. We have the same random factor as in the nullmodel: the mean intebsity may vary across persons. Additionally we now have a fixed factor to distinguish the two groups.

One-person-level-predictor model

This model allows you to assess if one predictor on the person level (i.e. one value for each grid, e.g. depression) can explain the average value of a constructs-level variable (e.g. intensity). In other words: can depression explain the average intensity score?

Example If construct level intensity scores are used as the dependent variable, you may assess if a person's depression score can predict a person's average level of intesity.

Select the variables first and then hit the run button.

Dependent variable

The dependent variable will usually be on the lowest hierarchical level (i.e. construct level).

Example We might want to understand how the construct intensity scores (which are nested within persons) variation can be explained.


Random variable

A random variable may vary randomly across the levels of another variable. For grids, usually the Intercept is allowed to vary across some other variable, e.g. the person or grid.


The variable

may vary across the levels of variable


Predictor

A metric predictor (i.e. a covariate) on the person level (e.g. depression score) that will explain the variation of the dependent variable.


The following variable is used as a metric predictor


Example We want to assess in how far the average intensity scores on constructs level can be explained by a person's' depression score.

One-person-level-predictor and one grouping factor

This model let's you evaluate if a predcitor can predcit an outcome on construct level and if there are differences in this relation differs across groups. We may e.g. want to know if construct intensity depends on the subjects depression score and if this relation is different for men and women.

Example If construct level intensity scores are used as the dependent variable, you may assess if a person's depression score can predict a person's average level of intesity.

Select the variables first and then hit the run button.

Dependent variable

The dependent variable will usually be on the lowest hierarchical level (i.e. construct level).

Example We might want to understand how the construct intensity scores (which are nested within persons) variation can be explained.


Random variable

A random variable may vary randomly across the levels of another variable. For grids, usually the Intercept is allowed to vary across some other variable, e.g. the person or grid.

The variable

may vary across the levels of variable


Predictor

A metric predictor (i.e. a covariate) on the person level (e.g. depression score) that will explain the variation of the dependent variable.

The following variable is used as a metric predictor

Example We want to assess in how far the average intensity scores on constructs level can be explained by a person's' depression score.


Group variable

A fixed effect, e.g. a factor to distinguish the control and treatment group.

The following variable distinguishes the groups

Example We want to assess if the intensity scores on constructs level differ between a treatment and a control group. We have the same random factor as in the nullmodel: the mean intebsity may vary across persons. Additionally we now have a fixed factor to distinguish the two groups.