I am interested to compare two groups i.e, credit access users vs non-users.
What steps I must follow?
psmatch2 credit_access (sex_hh age_hh family_size distance collateral attitude farm_size educ_recode marital_recode), kernel outcome( production_in_cmv ) bwidth(0.25) common logit ate
Mesfin Esayas The link shown below might be helpful to your cause.
https://www.statisticshowto.com/propensity-score-matching/
@Mesfin Esayas
"Some Practical Guidance for the Implementation of Propensity Score"
by M Caliendo can help you so that you can decide which matching algorithm to choose and determine (you can access the document easily).
I think you may reword your question: how to minimize biases using matching methods. Because matching does not control for unobservable characteristics that may cause selection bias, and as a result, the reliability of estimates is reduced.
The PSM method may produce estimates with low bias if datasets satisfy three conditions (Dehejia & Wahba, 2002): (i) data for treatment and control groups are collected using the same questionnaire; (ii) both treatment and control groups are drawn from the same locality; and (iii) the dataset contains a rich set of variables relevant to both treatment variable and the outcomes.
Next step is to choose between matching methods, e.g., kernel matching, radius matching, nearest neighbour matching, stratification matching...
Caliendo and Kopeinig (2008, p. 44) summarize trade-offs between bias and efficiency (low variance) for the matching estimators. No method outweighs, so choosing methods depends on the data structure at hand (Zhao, 2003). According to Zhao, the key data requirement for the PSM is that at each propensity score value or small score interval, the number of both treated and control observations is large enough to avoid the treated unit dropouts due to being not matched. In addition, if there is/are matches within a smaller (score) radius, the quality of matching will be improved, and matching estimates will be less biased. In the empirical studies, ATTK and ATTR are more commonly employed because the possibility of more control units used is higher for ATTK and ATTR relative to the other matching estimators, so they are more efficient. .
Hope this helps,
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments