Displayr

Displayr does not natively support many methods to allow a user to quantitatively evaluate the optimal number of clusters of a segmentation output. Suggest including native support for optimal cluster numbers like Elbow, Silhouette, and Gap score.

Currently, the tree output that a Latent Class Analysis generates has little to no options to format the look and feel for better readability on the screen. This can lead to cut-off variable labels or text that is too small to be read easily. This feature would bring many of the formatting options available in other outputs (font type, size, colors, etc.) to the LCA output.

In Displayr, add a function to save out respondent level segment membership probabilities from a Latent Class Analysis output similar to how it can be done in Q.

When updating an LCA segmentation with new data, it will break if there is a data structure change. However, it’s not always clear to the customer what has changed that will affect this. For example, changing the order of table categories and having the same non-missing data before and after are common issues, but this is not relayed in the error message.

The request is to increase the menu size and width for selecting variables for LCA in Displayr, so you can more clearly so what each variable is before selecting them:

There is confusion in the latent class segmentation tree output where the sizes shown on the tree are actually estimates based on the modeling and many times do not match the size of those segments in the variable created from the tree exactly. We have some more discussion on this on our technical documentation for trees here: https://wiki.q-researchsoftware.com/wiki/How_to_Interpret_Trees . The request is for the size to be relabeled in the output to reflect this.

In Q, user’s can click and drag the branches of an LCA output to move them around. But users cannot do the same in Displayr.