<style> .reveal { font-family: optima, noto sans tc; font-size: 30px; } .reveal .slides { text-align: left; } </style> <!-- .slide: style="text-align: center;" --> ## brief Chiu2012 :::success Chiu, M. S. (2012). The internal/external frame of reference model, big-fish-little-pond effect, and combined model for mathematics and science. *Journal of Educational Psychology, 104*(1), 87-107. ::: **Jack Tsai** 2021/4/6 <!-- Put the link to this slide here so people can follow --> slide: https://hackmd.io/@nccudoc/SJEtRimQ_ --- ### Intro (1/4) Internal/External of Refernce (I/E) Model - academic self-concept -> SC in diff ==domains== (e.g. math, science, English) - ==Positive effect==: Achievement and self-concept are matching. (internal comparison) - MAch(+) -> MSC(+) - VAch(+) -> VSC(+) - ==Negtive effect==: Achievement and self-concept are non-matching. (external comparison) - MAch(+) -> VSC(-) - VAch(+) -> MSC(-) --- ### Intro (2/4) Big-Fish-Little-Pond Efeect (BFLPE) - focused on academic self-concept either in general or particular domian. - students: - achieving group (+) -> self-concept (-) - achieving group (-) -> self-concept (+) --- ### Intro (3/4) Combine model - Internal comparison (individual lv) (I/E model) - Ach <-> SC - External comparison (group lv) (BFLPE) - self <-> group average |I/E model|BFLPE| |:--|:--| |1. multi-dim academic self-concept </br> 2. ==high== corr b/w Ach, ==low== corr b/w SC (non-matching) </br> 3. social comparison theory | 1. multi-dim academic self-concept </br> 2. social comparison theory | indivisual lv | group lv | --- ### Intro (4/4) Example ||| |:--:|:--:| ||Marsh & Hau (2004)| ||Chiu (2008)| ||Chiu (2012)| --- ### Method (1/2) #### Sample - student lv: 139,174 - school lv: 4,231 - country lv: 27 #### Measures - student lv: L1MAch, L1SAch, L1MSC, L1SSC - school lv: L2MAch, L2SAch - country lv: L3MAch, L3SAch --- ### Method (2/2) Statistical Analysis  #### random intercept model - only interpret the intercept (the level of `L1MSC`) - do not consider the interaction effect b/w variables - do not consider other interpretations on slope level. --- ### Results (1/2) math self-concept - Hypo 1/ **Model 2**: L1MSC <- ==(+)L1MAch== + ==(-)L1SAch== + SSC - Hypo 3/ **Model 4**: L1MSC <- ==(-)L2MAch== + L1MAch - Hypo 5/ **Model 7**: L1MSC <- ==(+)L1MAch== + ==(-)L1SAch== + ==(-)L2MAch== + L1SSC |L1MSC ~ .|model 2|model 4| model 7 | |:--|:--:|:--:|:--:| |(fixed)|||| |L1MAch|==.81***==|.80***|==.86***== |L1SAch|==-.17***==||==-.14***== |L1SSC|.09**||.08* |L2MAch||==-.50***==|==-.45***== |L2SAch||| |L3MAch||| |(random)||| |L1|.73|.74|.73 |L2|.09|.05|.05 |L3|.36|.11|.10 --- ### Results (2/2) science self-concept - Hypo 2/ **Model 10**: L1SSC <- ==(+)L1SAch== + ==(-)L1MAch== + MSC - Hypo 4/ **Model 12**: L1SSC <- ==(-)L2SAch== + L1SAch - Hypo 6/ **Model 15**: L1SSC <- ==(+)L1SAch== + ==(-)L1MAch== + ==(-)L2SAch== + L1MSC |L1SSC ~ .|model 10|model 12| model 15 | |:--|:--:|:--:|:--:| |(fixed)|||| |L1MAch|==-.09*==||==-.07== |L1SAch|==.52***==|.54***|==.54***== |L1MSC|.09**||.09* |L2MAch||| |L2SAch||==-.28***==|==-.24***== |L3MAch||| |(random)||| |L1|.77|.77|.76 |L2|.09|.07|.07 |L3|.23|.13|.11 --- ### Discussion (1/1) combined model 1) Individuals' self-concept in a domain is influenced by comparing their performance in other domains. 2) The BFLPE precisely captured the essence of social comparison theory. --- ### Question 1) Why multilevel researches usually present so many model results? (even if they only need a few of these models to test their hypo?) 2) In this paper, the author only set random effects on intercept (i.e., mean of math/science self-concept). If L2Ach and L3Ach are only explained to L1SC, why does it appear in the fixed effect?(what do they mean?) I am a little confused? 3) What kind of situation should we use random slope effect or interaction effect to interpret data?