Scatters of Factor Scores

[andkov]

The graph plots factor scores produced by Mplus and stored in .gh5 files. It matches the data in .gh5 to the parameters reports in .out files. Please help interpret the plots and spot oddities by submitting an entry into a table of interpretations. One cell corresponds to one graph. The graphs could be viewed grouped by process pairs for females and males.

Graph legend:

• gamma_00 / gamma_10 , semi-transparent grey lines: estimated fixed effects
• dashed line with transparent ribbon : loess smoother of factor scores
• straight red line: linear smoother of the factor scores
• red equation in bottom left: equation of the linear smoother
• black text in top right: Correlation b/w random terms (standard error), p-value

NOTE: some of the models are named inconsistently with the other. They are not mismatched, they just have their coordinates flipped because of the inconsistent naming. You can view the status of and comment to each model in the list of the submitted models.

[wibeasley]

@andkov

1. Within a panel, should the black R value correspond to the red r^2 value?
2. Consider increasing the span of the loess smoother, so it's not chasing those high leverage points

3. For the black equation, try making this change. It moves to the unused space in the top left, and also makes the alignment easier with Inf/-Inf. It can't find one of the gh5 files, so I can't run the graph myself.

   Try changing

   annotate(geom="text", size=baseSize-6, x=max_i2, y=max_i1, label=R_IPIC_display, hjust=1)+

   to

   annotate(geom="text", size=baseSize-6, x=-Inf, y=Inf, label=R_IPIC_display, hjust=1.1, vjust=-.1)+

[wibeasley]

Right now, the legends are determined for each graph separately. For instance the dark purple in the top legend means something different in the bottom legend. Consider if you want a colors to mean the same thing across studies & variable combinations. You may not... especially if these graph go to different publications and are never seen next to each other, or the BAGE meaning floats between different studies.

[andkov]

@wibeasley ,

1. No, red and black are different thing. Red is a linear smoother through the space of estimated factor scores. The black is a correlation coefficient taken from the output. They should line up and tell the same story ( if they both are ok indices). I've replaced R^2 to R for easier comparison.
2. Agree. I've changed it to span=1.5, although I don't quite understand the interpretation of the numeric value.
3. Agree. Implemented. Now the locations of the annotations are defined by Infs, vjust, and hjust
4. Color. You are right, for this case it would make sense to enforce the color values to the same scale across studies. However, I was thinking of using it in different installations: think of the graphs in this issue as elemental in the complex superplot, which would compare process pairs across studies.

[wibeasley]

think of the graphs in this issue as elemental in the complex superplot

When I work with you, @andkov, I think (& expect) of nothing less than a superplot.