SIMS 前後測配對 t 檢定報告

結論與樣本

在 35 筆可配對作答中，內在動機由 5.14 上升至 5.57，差異為 +0.44 分；經五項檢定的 Holm 校正後，仍達統計顯著，t(34) = 3.07，p_Holm = .021，d_z = 0.52。其餘四項分數在 Holm 校正後均未達顯著。

可配對作答35

前測未配對4

後測未配對6

配對資料遺漏題數0

主要發現：本資料支持「內在動機提升」的配對差異；其他構面的方向性變化可作為後續觀察，但不是本次校正後的統計證據。

配對與分析方法

以兩份工作簿「答題記錄(完整選項內容)」工作表的受試者識別碼進行配對，前測 39 份、後測 41 份，共得到 35 筆完整配對。
使用前置的 1–7 分數解讀 Likert 方向；內在動機為 Q1/Q5/Q9/Q13，認同調節為 Q2/Q6/Q10/Q14，外在調節為 Q3/Q7/Q11/Q15，無動機為 Q4/Q8/Q12/Q16。
每一構面以 4 題平均分數計算。整體正向動機合成分數僅作補充描述，將無動機題 Q4/Q8/Q12/Q16 反向計分後納入 16 題平均；它不是獨立的 SIMS 構面。
所有檢定為雙尾配對 t 檢定，差異定義為「後測 − 前測」。同時檢定五項分數，p 值以 Holm 法校正；效果量採配對設計的 Cohen’s d_z。

配對 t 檢定結果

n = 35；df = 34；差異 = 後測 − 前測

分數	前測 M	後測 M	差異	t	原始 p	Holm p	d_z	95% CI（差異）	校正後
整體正向動機合成分數	5.19	5.45	+0.26	2.60	.014	.055	0.44	[0.06, 0.46]	未達顯著
內在動機	5.14	5.57	+0.44	3.07	.004	.021	0.52	[0.15, 0.72]	顯著
認同調節	5.90	6.07	+0.17	1.40	.171	.341	0.24	[−0.08, 0.42]	未達顯著
外在調節	5.14	5.28	+0.14	1.09	.284	.341	0.18	[−0.12, 0.41]	未達顯著
無動機	3.41	3.12	−0.29	−1.97	.057	.171	−0.33	[−0.58, 0.01]	未達顯著

結果解讀

內在動機：校正後仍達顯著，效果量 d_z = 0.52，屬中等幅度。95% 信賴區間完全高於 0，顯示在可配對樣本中，後測平均分數較高。
整體正向動機合成分數：原始 p = .014，但 Holm 校正後為 .055，剛未達 .05；此結果不應寫成校正後顯著。
認同調節與外在調節：平均數為正向差異，但信賴區間包含 0，證據不足以判定實質改變。
無動機：平均數下降 0.29 分，方向上是有利的，然而 p_Holm = .171，尚不足以作出顯著下降的結論。

限制與建議

結果只反映 35 筆可配對作答；未配對的前測 4 筆與後測 6 筆不納入 t 檢定。
配對 t 檢定支持配對資料中的平均差異，但未納入對照組、隨機分派或其他外在因素，因此不宜單憑本結果做強因果主張。
樣本數低於 16 題的五倍門檻（80），本報告不以 KMO、Bartlett、EFA 或 CFA 當作穩定的建構效度證據。
下一步可預先指定內在動機為主要結果、報告效果量與信賴區間，並持續收集更大樣本以檢驗其可重現性。

Summary and sample

Across 35 matched records, intrinsic motivation increased from 5.14 to 5.57 (+0.44). It remained statistically significant after Holm correction across five tests: t(34) = 3.07, p_Holm = .021, d_z = 0.52. The other four scores were not significant after correction.

Matched responses35

Unmatched pretest records4

Unmatched posttest records6

Missing paired item values0

Main finding: The paired data support an increase in intrinsic motivation. Directional changes in the other subscales can be monitored, but they are not corrected statistical evidence in this analysis.

Pairing and method

Records were paired by respondent identifier in the full-text response-record worksheets: 39 pretest records and 41 posttest records yielded 35 complete pairs.
The leading 1–7 values were used for Likert direction. Intrinsic motivation: Q1/Q5/Q9/Q13; identified regulation: Q2/Q6/Q10/Q14; external regulation: Q3/Q7/Q11/Q15; amotivation: Q4/Q8/Q12/Q16.
Each subscale is the mean of its four items. The overall positive-motivation composite is supplementary only: the four amotivation items were reverse-scored before all 16 items were averaged; it is not an independent SIMS subscale.
All tests are two-sided paired t-tests, defining change as posttest minus pretest. Five tests were adjusted with Holm’s procedure. Effect size is paired-design Cohen’s d_z.

Paired t-test results

n = 35; df = 34; change = posttest − pretest

Score	Pre M	Post M	Change	t	Raw p	Holm p	d_z	95% CI (change)	Adjusted result
Overall positive-motivation composite	5.19	5.45	+0.26	2.60	.014	.055	0.44	[0.06, 0.46]	Not significant
Intrinsic motivation	5.14	5.57	+0.44	3.07	.004	.021	0.52	[0.15, 0.72]	Significant
Identified regulation	5.90	6.07	+0.17	1.40	.171	.341	0.24	[−0.08, 0.42]	Not significant
External regulation	5.14	5.28	+0.14	1.09	.284	.341	0.18	[−0.12, 0.41]	Not significant
Amotivation	3.41	3.12	−0.29	−1.97	.057	.171	−0.33	[−0.58, 0.01]	Not significant

Interpretation

Intrinsic motivation: The corrected result is significant with a medium effect (d_z = 0.52). Its 95% confidence interval is entirely above zero.
Overall positive-motivation composite: Raw p = .014, but Holm-adjusted p = .055. It must not be described as significant after correction.
Identified and external regulation: Mean changes are positive, but the confidence intervals include zero; the evidence is insufficient to conclude a reliable change.
Amotivation: The mean decreased by 0.29 points, a favorable direction, but p_Holm = .171 does not support a significant decrease.

Limits and next steps

Results use only 35 matched records; four pretest and six posttest records were not included in the paired tests.
Paired t-tests support mean differences in the matched data, but there is no control group, random assignment, or adjustment for external influences. Strong causal claims are not warranted from this result alone.
The sample is below five times the 16-item count (80), so KMO, Bartlett, EFA, and CFA are not presented as stable construct-validity evidence.
Next work can pre-specify intrinsic motivation as the primary outcome, report effect sizes and confidence intervals, and add a larger sample to test reproducibility.