[問題] Logistic 如何跑 Joint Test？
我的 DV 是有調節作用的 Logistic, 評審要我再報告出交互作用的 "Joint test"
我查了一下，所謂 Joint test 似乎不止是作圖
我找了一些有做「Joint test」的 Paper，
似乎是要看「兩個迴歸模型」之間的係數差異是否顯著
如果是這樣，在一般連續型迴歸的軟體裡似乎是看眾模式間的 R2 change
問題是 Logistic 並沒有這個選項
故想問，Logistic 該怎麼跑 Joint Test呢？
────────────────────────────────────
註：我找了一些有做 Joint Test的研究，其 Result 大部分是這樣寫：
例一：We perform a statistical test of whether the effect of a patent grant
varies by technology area;
the joint test of all interaction terms equal zero gives an F test statistic
of 1.3, which has a p-value of 0.142. Thus we find no evidence to suggest that
the effect of a patent grant is different across technology areas.
例二：A joint test of significance also indicated that four main outside
director variables were significantly different than zero (F = 4.55, p < .034)
例三：A joint test for equality of means is rejected with a p-value of 0.025, while
a one-sided test of no complementarity—
i.e., testing the incremental effect of adding an innovation activity—is
rejected at a 5% level of significance.
例四：In Hypothesis 3 we predicted that acquisition experience would positively
moderate the effect of structural integration on knowledge leverage.
From Model 3a in Table 3 we see that the coefficient for the interaction term
is negative but not significant.
We also conducted a joint test of significance for structural integration and
the interaction term to assess whether collinearity between these terms was
suppressing a significant effect.
However, we found no evidence for this as the coefficients are not jointly
significant (χ2(2) = 3.48, p = 0.17). Hence, Hypothesis 3 is not supported.
例五：The third estimation is the same as the second, except that the sample
excludes acquirers that made more than one acquisition in the time period.
All variables are robustly estimated, and a joint test on all the
coefficients in the third regression cannot reject the null hypothesis that
they equal those found in the regression with the full sample (χ2(11)=3.52,
prob>χ2=0.9819).
Domestic acquirers that made only one acquisition have the same decision
patterns as those that made more than one acquisition
The results in the third column of Table 6 show that the estimation in column
2 regressed on the subsample of acquirers that made only one acquisition
during the time period contributes positively to the model’s overall fit.
However, a joint test on all the coefficients in the third regression cannot
reject the null hypothesis that they equal those found in the regression with
the full sample (χ2(11)=2.80, prob>χ2=0.9932).
It appears that the decision process of foreign acquirers that made only one
acquisition does not differ from those that made more than one acquisition.
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 140.114.139.134
→
12/25 22:53, , 1F
12/25 22:53, 1F
→
12/25 22:54, , 2F
12/25 22:54, 2F
→
12/25 23:38, , 3F
12/25 23:38, 3F