SPSS Statistics

 View Only

Introducing two new analytical macros to SPSS users

  • 1.  Introducing two new analytical macros to SPSS users

    Posted Wed May 31, 2023 11:16 AM
    Edited by Kirill Orlov Wed May 31, 2023 11:47 AM

    Two new large macros have been recently published on Kirill's SPSS Macros Page https://www.spsstools.net/en/KO-spssmacros .

    If the Group kindly permit me, I'll inform of that here.

    1) !KO_AMONGCATS macro

    Compares proportions of the categories defined by one categorical variable ("single response" variable) or proportions of the positive response in a set of binary variables ("multiple response" set). Results output is formatted similar to Custom Tables, and the procedure is apt for processing survey data.

    SPSS Custom Tables currently does pairwise statistical comparisons of proportions only between independent groups, for example, to compare the proportion of a "Yes" response between groups "Females" vs "Males" vs "Other". !KO_AMONGCATS will compare between the categories' proportions themselves, i.e., compare the sizes of groups "Females" vs "Males" vs "Other". It is the "one-sample comparison". Or, if the "categories" are not exclusive, such as after a multiple-response question ("Which brands do you know? "Audi", "Daewoo",  "Ford", etc.), !KO_AMONGCATS will compare proportions of positive response (selection rate) between those. It is the "paired-samples comparison".

    Some features of the macro:
    - Exact and asymptotic testing
    - Optional omnibus test before pairwise testing
    - Bonferroni and Benjamini-Hochberg multiple comparison corrections
    - Simple and APA-subscript style of display of significant results (as in Custom Tables)
    - Optional detailed printout
    - Plotting a graph (network etc.) to see significant differences

    The macro positions itself as a potentially valuable complement for Custom Tables in survey data analysis in social sciences, marketing research etc.

    2) !KO_SEQSIM macro

    Compares sequences of elements with computation of a pairwise similarity measure between the sequences. In a comparison, one may take or not take into account chains of elements or their location – i.e., compare as sequences or simply as documents. The similarity is calculated either through matching method or through aligning method.

    Some features of the macro:
    - Consider or ignore order of adjacent "words" (elements) while comparing
    - Consider or ignore general location within the body while comparing
    - Tune how much longer common chains influence the similarity than shorter common chains
    - Set degree of relevance of "words" or their synonymy/closeness
    - Imitate search query by identifying longest common subsequence.

    There generally exist many methods and algorithms to compare documents and compare sequences. They are of different ideologies and applications. The approaches on which !KO_SEQSIM is based are more frequently employed in social studies than, put it, in molecular biology. You may well use the macro in text analysis (after appropriate preparation of the data).

    Kirill Orlov