Internet speed plays the role in the communication with the server so it may affect any computation step in the MPI Central, in the indicators fetching and in the data upload.
Upload time may vary depending on internet upload speed at your network. For example in a home internet where the internet speed is 4 Mb per seconds, the upload speed will be 1 Mb per second. Fiber optic or cable connections will be optimal for fast upload speed.
Other factors affecting upload speed: The number of variables and number of records also affect the upload time. The larger the data file the longer it will take to be uploaded. On average 130,000 records with 100 variables should take approximately 3 minutes to upload.
Excel data files uploaded in the survey tab should not exceed a 100 MB, as the upload will be too slow. The average size (based on common surveys) is around 20 MB. In case your file is too large, we suggest deleting all string variables from the data and/or variables that are irrelevant for an MPI framework, when preparing the data for upload.
The larger the data file the longer it will take to be uploaded. On average 130,000 records with 100 variables should take approximately 3 minutes to upload.
Yes, we provide customized training sessions for officials in Member States, on demand. However, short online tutorials can be found here.
Different types of surveys can be uploaded on MAT. The tool is flexible and its algorithm cares mainly about the structure of the variables in the data files, where the MAT will recognize the record identification variables in the uploaded files.
Since the underlying computation method is the Alkire-Foster methodology, the minimum requirements is to upload the Household data file and the Individual data file. Within those files you should specify a standard name for the key variable that will link the records of the two files. Users should pay special attention to the survey design to avoid un-necessary missingness or lack of representativity of results (see How to upload my data correctly)
Within the Wizard, MAT is case sensitive. However, when uploading the data, the variables names are not differentiated by capitalization. In other words, if you have 2 age variables in your data set "Age" and "age" MAT will not differentiate them at the computation level (see section on percentage of non-missingness > 100%). in this case, MAT will prompt a warning for duplicate variables.
A. Users need to distinguish unique names for variables. The MAT is not case sensitive. If two variables have the same name with one capitalized letter difference, the data structure in MAT will count observations in both variables.
B. Whenever the percentage of data availability is greater than a 100%, there is something wrong in the template file, mainly in identifying the line numbers.
(i) If the line number is not well defined as a variable (see Template User Manual), MAT will only identify unique records but will count all responses, thus the number of populated answers will be higher than the total identified records.
(ii) The records distinction by MAT is done at the Cluster ID (CL), Household ID (HH), Individual ID (LN) level. The individual ID has to be unique in every file except in the Birth History type of files where the woman ID can be repetitive for each birth she ever gave; MAT is designed to recognize this structure when uploading a data file in the Birth History field, only. MAT will not verify for duplication, it is the responsibility of the User .
For them to show correctly in the maps on MAT, the names of the governorates in the uploaded data must be identical to the list found in the template user manual. If there is a mismatch in the names (data and map reference list), the governorate results will not show.
To be recognized in MAT, the type of variables must be:
The variable cannot be: Integer
In steps 1 to 5 (Select Survey, Define Dimensions, Deprivation Cut-offs, Set Weights, MPI Calculation) in the MPI Central Tab, MAT works at level of raw observations or records. Only in Step 5, when the user clicks "Calculate MPI", the algorithm will expand the individual deprivation status to the household level, based on the indicatorâ€™s definition and the relevant cut-off points.
So, any changes in the indicator cut-off will be applied to the raw data and transform it to binary. Then this change will be expanded to the rest of the HH members following the indicator specification (ALL/ANY).
There is no limit to the number of indicators per dimension, in the MPI Central tab, nor on the number of dimensions to be used by the user. However, based on the principle of parsimony, we recommend using a minimum number of indicators that would proxy the latent dimension to be measure. In practice, we recommend not to go beyond 5 indicators per dimension and not more than 7 dimensions per framework.
The average fetching time is about 5 seconds, please give it another few seconds if your indicator has many variables (and/or operator etc). If you are experiencing a fetching time longer than 1 minute, please check your internet connection or kindly contact the administrator on firstname.lastname@example.org
There are two types of errors in the fetching process:
1- Pop-up notification that the variables used to build the indicator no longer exist, which means that they cannot be found in the data (deleted or renamed); or
2- When you click the fetch button the deprivation results show you zeros (check image below), then the indicatorâ€™s rules could be empty. Otherwise, kindly contact the administrator on email@example.com
Several input can be changed in a framework: